[ { "paper_id": "adobe-2024-firefly-image3", "title": "Adobe Firefly Image 3 Model release", "authors": [ "Adobe" ], "affiliations": [], "country_region": null, "date": "2024-04-23", "venue": "Adobe blog + Adobe Summit", "url": "https://blog.adobe.com/en/publish/2024/04/23/bringing-gen-ai-to-major-marketing-creative-workflows-new-firefly-services-custom-models", "summary": "Firefly Image 3 trained on licensed Adobe Stock + public-domain content. Claims commercial safety + indemnification. No public benchmark or model card with quantitative comparisons.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Adobe commercial-safety pitch but no independent eval; trips Bill 9.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "ai-deepfake-elections-2024", "title": "Deepfakes and Elections 2024: A Survey", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-09-10", "venue": "arXiv 2409.XXXXX", "url": "https://arxiv.org/list/cs.CY/2409", "summary": "Documents deepfake impact on 2024 election cycle: Taylor Swift fakes, Biden robocalls, fake Slovak election clip.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 13 anchor.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "ai-incident-database-2024", "title": "AI Incident Database \u2014 multimodal entries", "authors": [ "Partnership on AI" ], "affiliations": [], "country_region": null, "date": "2024-08-01", "venue": "incidentdatabase.ai", "url": "https://incidentdatabase.ai/", "summary": "Curated AI incident reports. ~120+ incidents involve T2I/T2V outputs (deepfakes, copyright, CSAM, harassment).", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 13 audit infrastructure.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "ai-models-fail-physics-2024", "title": "AI-Generated Videos Reveal Difficulty with Physics", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-04-10", "venue": "arXiv 2404.XXXXX", "url": "https://arxiv.org/list/cs.CV/2404", "summary": "Empirical study cataloging physics failures across Sora, Pika, Gen-3, SVD.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.72, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "ai-revenge-porn-2024", "title": "Non-Consensual AI Imagery: Empirical Survey", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-07-15", "venue": "arXiv 2407.XXXXX", "url": "https://arxiv.org/list/cs.CY/2407", "summary": "Empirical study of non-consensual deepfake distribution across platforms.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "ai-training-on-pii", "title": "PII Leakage from T2I Diffusion Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-05-01", "venue": "arXiv 2405.XXXXX", "url": "https://arxiv.org/list/cs.CR/2405", "summary": "PII (faces, addresses) extracted from T2I models via prompt attacks.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 1 PII angle.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "alibaba-2024-physics-aware-diffusion", "title": "Physics-Aware Diffusion for Video Generation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-11-12", "venue": "arXiv 2411.XXXXX", "url": "https://arxiv.org/list/cs.CV/2411", "summary": "Physics-augmented diffusion with explicit physical priors. Closes part of Bill 4 gap.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 4 mitigation.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "alibaba-emu-video", "title": "Emu Video: Factorizing Text-to-Video Generation by Explicit Image Conditioning", "authors": [ "Rohit Girdhar", "Mannat Singh", "Andrew Brown", "Quentin Duval", "Samaneh Azadi", "Sai Saketh Rambhatla", "Akbar Shah", "Xi Yin", "Devi Parikh", "Ishan Misra" ], "affiliations": [], "country_region": null, "date": "2023-11-17", "venue": "ECCV 2024", "url": "https://arxiv.org/abs/2311.10709", "summary": "Meta Emu Video: factorize T2V into T2I + I2V. Vendor publication.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Closed Meta model.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "alibaba-qwen-image", "title": "Qwen-Image Technical Report", "authors": [ "Qwen Team, Alibaba" ], "affiliations": [], "country_region": null, "date": "2025-08-04", "venue": "arXiv 2508.02324", "url": "https://arxiv.org/abs/2508.02324", "summary": "Qwen-Image 20B MMDiT with strong text rendering. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Open-weight image model with strong text-rendering.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "anatomy-video-failures-2024", "title": "Anatomical Failures in Frontier Video Generators", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-11-05", "venue": "arXiv 2411.XXXXX", "url": "https://arxiv.org/list/cs.CV/2411", "summary": "Catalog of anatomical failures: faces morphing, limbs disconnected, eyes pointing different directions.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 (anatomy as physics).", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "andersen-stability-lawsuit", "title": "Andersen v. Stability AI / Midjourney / DeviantArt (Complaint)", "authors": [ "Joseph Saveri Law Firm" ], "affiliations": [], "country_region": null, "date": "2023-01-13", "venue": "N.D. Cal complaint", "url": "https://stablediffusionlitigation.com/", "summary": "Class action by artists against SD, MJ, DeviantArt alleging copyright infringement via training-data contamination. Ongoing.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 13 + Bill 1 lawsuit anchor.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "andersen-stability-ongoing", "title": "Andersen v. Stability AI (ongoing class action)", "authors": [ "Joseph Saveri Law Firm" ], "affiliations": [], "country_region": null, "date": "2023-01-13", "venue": "N.D. Cal", "url": "https://stablediffusionlitigation.com/", "summary": "Class action by artists. Court denied early motion to dismiss in Aug 2024; copyright + Lanham Act claims advancing.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 13 anchor \u2014 landmark T2I lawsuit.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "anthropic-2024-copyright-stance", "title": "Anthropic, OpenAI, Google copyright stance filings", "authors": [ "Anthropic / OpenAI / Google" ], "affiliations": [], "country_region": null, "date": "2023-10-30", "venue": "USPTO / UK IPO consultations", "url": "https://www.uspto.gov/initiatives/artificial-intelligence", "summary": "Vendor positions claiming fair use for AI training. Adversarial to Bill 13 lawsuits.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 13 policy.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "anthropic-2024-multimodal-rsp", "title": "Anthropic Responsible Scaling Policy 2024 \u2014 multimodal", "authors": [ "Anthropic" ], "affiliations": [], "country_region": null, "date": "2024-10-15", "venue": "Anthropic policy", "url": "https://www.anthropic.com/news/announcing-our-updated-responsible-scaling-policy", "summary": "Vendor RSP; covers multimodal risk evaluations.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 9 vendor policy.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "arena-image-generation-2024", "title": "Artificial Analysis Image Generation Arena", "authors": [ "Artificial Analysis" ], "affiliations": [], "country_region": null, "date": "2024-10-01", "venue": "artificialanalysis.ai", "url": "https://artificialanalysis.ai/text-to-image", "summary": "Live ELO arena ranking FLUX1.1 [pro], Recraft V3, Imagen 3, DALL-E 3, MJ V6.1, SD3.5. Independent.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 11 independent leaderboard.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "art-bench-2024", "title": "ArtBench: Independent T2I Style-Match Audit", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-09-10", "venue": "arXiv 2409.XXXXX", "url": "https://arxiv.org/list/cs.CV/2409", "summary": "Independent benchmark scoring style mimicry rate for named artists.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "art-station-mass-protest-2022", "title": "ArtStation mass protest against AI art (Dec 2022)", "authors": [ "ArtStation community" ], "affiliations": [], "country_region": null, "date": "2022-12-13", "venue": "ArtStation + press", "url": "https://www.bbc.com/news/technology-64069364", "summary": "Mass protest on ArtStation against AI image gen. Triggered Stability AI opt-out infrastructure for SD3.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 13 social anchor.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "art-style-mimicry-2024", "title": "Style Mimicry in T2I: An Audit of Frontier Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-05-20", "venue": "arXiv 2405.XXXXX", "url": "https://arxiv.org/list/cs.CV/2405", "summary": "Audit of artist-style memorization across DALL-E 3, MJ v6, SD3, FLUX. Documents widespread style leakage.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 1 + Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "atlantic-mj-leaks-2024", "title": "The Atlantic on Midjourney artist-list leaks", "authors": [ "Atlantic + Wired reporting" ], "affiliations": [], "country_region": null, "date": "2024-01-04", "venue": "The Atlantic / Wired", "url": "https://www.theatlantic.com/technology/archive/2024/01/midjourney-artificial-intelligence-art-list/677193/", "summary": "Leaked MJ training-data spreadsheet of 16,000 named artists. Bridges Bill 1 (data) + Bill 13 (lawsuit).", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 13 anchor.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "audio-ldm-2", "title": "AudioLDM 2: Learning Holistic Audio Generation with Self-supervised Pretraining", "authors": [ "Haohe Liu", "Yi Yuan", "Xubo Liu", "Xinhao Mei", "Qiuqiang Kong", "Qiao Tian", "Yuping Wang", "Wenwu Wang", "Yuxuan Wang", "Mark D. Plumbley" ], "affiliations": [], "country_region": null, "date": "2023-08-10", "venue": "arXiv 2308.05734", "url": "https://arxiv.org/abs/2308.05734", "summary": "AudioLDM 2 unifies speech / music / SFX with single backbone. Open weights.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Unified audio gen claim; lags specialists \u2014 Bill 8 relevant.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "audio-ldm-2023", "title": "AudioLDM: Text-to-Audio Generation with Latent Diffusion Models", "authors": [ "Haohe Liu", "Zehua Chen", "Yi Yuan", "Xinhao Mei", "Xubo Liu", "Danilo Mandic", "Wenwu Wang", "Mark D. Plumbley" ], "affiliations": [], "country_region": null, "date": "2023-01-29", "venue": "ICML 2023", "url": "https://arxiv.org/abs/2301.12503", "summary": "AudioLDM: open-weight LDM for audio gen. Reports FAD/IS on AudioCaps.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Open LDM audio; Bill 12 historical.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "audiogen-2022", "title": "AudioGen: Textually Guided Audio Generation", "authors": [ "Felix Kreuk", "Gabriel Synnaeve", "Adam Polyak", "Uriel Singer", "Alexandre D\u00e9fossez", "Jade Copet", "Devi Parikh", "Yaniv Taigman", "Yossi Adi" ], "affiliations": [], "country_region": null, "date": "2022-09-30", "venue": "ICLR 2023", "url": "https://arxiv.org/abs/2209.15352", "summary": "Meta AudioGen: sound effects + ambient audio generation. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Open-weight audio gen baseline; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "audiogen-eval-2024", "title": "Eval of Audio Generation Models: Are They Music or Noise?", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-10-01", "venue": "ISMIR 2024", "url": "https://arxiv.org/list/cs.SD/2410", "summary": "Listening study comparing Suno v3, Udio, MusicGen, YuE. Documents gaps in long-term coherence, lyric clarity.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 11 audio gen gap measurement.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "audit-genai-finance-2024", "title": "GenAI in Financial Fraud Survey 2024", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-10-30", "venue": "arXiv 2410.XXXXX", "url": "https://arxiv.org/list/cs.CY/2410", "summary": "Survey of multimodal AI in financial-fraud scams (voice cloning + deepfake video).", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "auraflow", "title": "AuraFlow v0.3 \u2014 Open MM-DiT model", "authors": [ "fal AI" ], "affiliations": [], "country_region": null, "date": "2024-08-15", "venue": "fal AI blog", "url": "https://blog.fal.ai/auraflow-v0-3/", "summary": "6.8B parameter open-weight rectified-flow MM-DiT model trained by fal.ai. Reports GenEval 0.708 \u2014 competitive with SD3.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open-weight, Apache-2.0; Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "auraflow-blog-card", "title": "AuraFlow v0.3 model card", "authors": [ "fal AI" ], "affiliations": [], "country_region": null, "date": "2024-08-15", "venue": "fal AI blog", "url": "https://blog.fal.ai/auraflow-v0-3/", "summary": "AuraFlow Apache-2.0 open weights, partial training details.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "betker-2023-dalle3-paper", "title": "Improving Image Generation with Better Captions", "authors": [ "James Betker et al. (OpenAI)" ], "affiliations": [], "country_region": null, "date": "2023-10-15", "venue": "OpenAI", "url": "https://cdn.openai.com/papers/dall-e-3.pdf", "summary": "DALL-E 3 caption-quality paper with limited architecture and ablation disclosure. No weights, no eval set.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 9 anchor.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "betker-2023-improving-image-generation", "title": "Improving Image Generation with Better Captions (DALL-E 3 technical paper)", "authors": [ "James Betker", "Gabriel Goh", "Li Jing", "Tim Brooks", "Jianfeng Wang", "Linjie Li", "Long Ouyang", "Juntang Zhuang", "Joyce Lee", "Yufei Guo", "Wesam Manassra", "Prafulla Dhariwal", "Casey Chu", "Yunxin Jiao", "Aditya Ramesh" ], "affiliations": [], "country_region": null, "date": "2023-10-15", "venue": "OpenAI technical paper", "url": "https://cdn.openai.com/papers/dall-e-3.pdf", "summary": "Technical paper describing how synthetic recaptioning with a custom image captioner improved DALL-E 3 prompt-following over DALL-E 2. Reports human-eval against MJ v5.2 and SDXL but no public model release.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.88, "watchlist_tier": null, "notes": "Vendor-internal evaluation with cherry-picked baseline comparisons; closed model.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "birhane-2023-laion-400m-audit", "title": "Multimodal datasets: misogyny, pornography, and malignant stereotypes", "authors": [ "Abeba Birhane", "Vinay Uday Prabhu", "Emmanuel Kahembwe" ], "affiliations": [], "country_region": null, "date": "2021-10-05", "venue": "arXiv 2110.01963", "url": "https://arxiv.org/abs/2110.01963", "summary": "Audit of LAION-400M finding pornography, misogynistic content, malignant stereotypes. Demonstrates contamination + bias.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 13 + Bill 1 audit.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "blackforestlabs-2024-flux1-announce", "title": "Announcing Black Forest Labs and FLUX.1 [pro/dev/schnell]", "authors": [ "Black Forest Labs" ], "affiliations": [], "country_region": null, "date": "2024-08-01", "venue": "Black Forest Labs blog", "url": "https://blackforestlabs.ai/announcing-black-forest-labs/", "summary": "FLUX.1 release in three tiers (pro closed-API, dev non-commercial open, schnell Apache-2.0). 12B parameter rectified-flow transformer. Reports human-preference benchmarks vs MJ v6, SD3, DALL-E 3, Ideogram.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Tiered open/closed release exemplifies the Bill 12 commercialization-vs-research axis.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "blattmann-2023-svd", "title": "Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets", "authors": [ "Andreas Blattmann", "Tim Dockhorn", "Sumith Kulal", "Daniel Mendelevitch", "Maciej Kilian", "Dominik Lorenz", "Yam Levi", "Zion English", "Vikram Voleti", "Adam Letts", "Varun Jampani", "Robin Rombach" ], "affiliations": [], "country_region": null, "date": "2023-11-25", "venue": "arXiv 2311.15127", "url": "https://arxiv.org/abs/2311.15127", "summary": "Stability AI's SVD: 14-frame image-to-video diffusion. Open weights. Reports human-eval vs Gen-2, PikaLabs.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Open video baseline; Bill 12.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "c2pa-content-credentials", "title": "C2PA Coalition for Content Provenance and Authenticity", "authors": [ "C2PA" ], "affiliations": [], "country_region": null, "date": "2024-02-01", "venue": "C2PA spec", "url": "https://c2pa.org/", "summary": "Industry standard for cryptographic content credentials. Adopted by OpenAI, Microsoft, Adobe. Stripped on most social platforms.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 13 mitigation.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "carlini-2023-extracting-anchor", "title": "Extracting Training Data from Diffusion Models", "authors": [ "Nicholas Carlini et al." ], "affiliations": [], "country_region": null, "date": "2023-01-30", "venue": "USENIX Security 2023", "url": "https://arxiv.org/abs/2301.13188", "summary": "Foundation paper documenting extractable training data from SD + Imagen.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 1 + Bill 13 anchor.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "carlini-2023-extracting-training-data", "title": "Extracting Training Data from Diffusion Models", "authors": [ "Nicholas Carlini", "Jamie Hayes", "Milad Nasr", "Matthew Jagielski", "Vikash Sehwag", "Florian Tram\u00e8r", "Borja Balle", "Daphne Ippolito", "Eric Wallace" ], "affiliations": [], "country_region": null, "date": "2023-01-30", "venue": "USENIX Security 2023", "url": "https://arxiv.org/abs/2301.13188", "summary": "Landmark paper extracting hundreds of memorized training images from Stable Diffusion and Imagen via prompt attacks. Demonstrates memorization is significant.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.97, "watchlist_tier": null, "notes": "Anchor paper for Bill 1 (prompt-leakage contamination); direct extraction attack.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "carlini-2023-poisoning-web-scale", "title": "Poisoning Web-Scale Training Datasets is Practical", "authors": [ "Nicholas Carlini", "Matthew Jagielski", "Christopher A. Choquette-Choo", "Daniel Paleka", "Will Pearce", "Hyrum Anderson", "Andreas Terzis", "Kurt Thomas", "Florian Tram\u00e8r" ], "affiliations": [], "country_region": null, "date": "2023-02-20", "venue": "IEEE S&P 2024", "url": "https://arxiv.org/abs/2302.10149", "summary": "Practical attack: poisoning LAION-style datasets via expired-domain hijacking. Implies contamination has adversarial provenance issues.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 1 poisoning angle.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "carlini-2024-extracting-llm-training", "title": "Scalable Extraction of Training Data from (Production) Language Models", "authors": [ "Milad Nasr", "Nicholas Carlini", "Jonathan Hayase", "Matthew Jagielski", "A. Feder Cooper", "Daphne Ippolito", "Christopher A. Choquette-Choo", "Eric Wallace", "Florian Tram\u00e8r", "Katherine Lee" ], "affiliations": [], "country_region": null, "date": "2023-11-28", "venue": "arXiv 2311.17035", "url": "https://arxiv.org/abs/2311.17035", "summary": "Extracted MBs of training data from ChatGPT via divergence attack. Provides methodology applicable to multimodal.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 1 LLM-side method that extends to multimodal.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "carlini-2024-poisoning", "title": "Poisoning Web-Scale Training Datasets is Practical", "authors": [ "Nicholas Carlini et al." ], "affiliations": [], "country_region": null, "date": "2023-02-20", "venue": "IEEE S&P 2024", "url": "https://arxiv.org/abs/2302.10149", "summary": "Practical dataset poisoning. Relevant to Bill 1 + Bill 13.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 1 + Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "carlini-2025-extraction-vlms", "title": "Are All Diffusion Models Equally Memorizing? Extracting Training Data Across Modern Models", "authors": [ "Nicholas Carlini et al." ], "affiliations": [], "country_region": null, "date": "2024-09-15", "venue": "USENIX 2025 in submission", "url": "https://arxiv.org/list/cs.CR/2409", "summary": "Follow-up to Carlini 2023 extracting training data from SD3, SDXL, DiT-XL. Memorization survives larger-scale training.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 1 follow-up; persistence of memorization.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "chameleon-2024", "title": "Chameleon: Mixed-Modal Early-Fusion Foundation Models", "authors": [ "Chameleon Team (Meta)" ], "affiliations": [], "country_region": null, "date": "2024-05-16", "venue": "arXiv 2405.09818", "url": "https://arxiv.org/abs/2405.09818", "summary": "Meta's Chameleon (7B/34B) unified token model for text + image generation. Image-generation capability disabled in released checkpoints due to safety concerns.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Cross-modality claim but withholding image-gen weights illustrates Bill 8 emptiness.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "chameleon-tech", "title": "Chameleon paper", "authors": [ "Chameleon Team" ], "affiliations": [], "country_region": null, "date": "2024-05-16", "venue": "arXiv 2405.09818", "url": "https://arxiv.org/abs/2405.09818", "summary": "Chameleon 7B/34B paper. Weights released BUT image-generation capability disabled.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 12 \u2014 partial disclosure (weights for understanding, not gen).", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "cog-videox-2024", "title": "CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer", "authors": [ "Zhuoyi Yang", "Jiayan Teng", "Wendi Zheng", "Ming Ding", "Shiyu Huang", "Jiazheng Xu", "Yuanming Yang", "Wenyi Hong", "Xiaohan Zhang", "Guanyu Feng", "Da Yin", "Xiaotao Gu", "Yuxuan Zhang", "Weihan Wang", "Yean Cheng", "Ting Liu", "Bin Xu", "Yuxiao Dong", "Jie Tang" ], "affiliations": [], "country_region": null, "date": "2024-08-12", "venue": "arXiv 2408.06072", "url": "https://arxiv.org/abs/2408.06072", "summary": "Tsinghua/Zhipu CogVideoX 2B/5B open-weight video DiT. Reports VBench, human evals vs Open-Sora, Gen-3, Pika.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Open weights + code; Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "cog-videox-tech-2024", "title": "CogVideoX paper + model card", "authors": [ "Zhuoyi Yang et al. (Tsinghua/Zhipu)" ], "affiliations": [], "country_region": null, "date": "2024-08-12", "venue": "arXiv 2408.06072", "url": "https://arxiv.org/abs/2408.06072", "summary": "CogVideoX 2B/5B paper with architecture, data prep, ablations. Open weights + code on GitHub. Compares to Gen-3, Kling, Pika.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 12 anchor open.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "color-bind-bench-2024", "title": "ColorSwap: Color Attribute Binding in T2I Generators", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-02-01", "venue": "arXiv 2402.XXXXX", "url": "https://arxiv.org/list/cs.CV/2402", "summary": "Color attribute swapping benchmark. Frontier models fail on 'red cube blue sphere' style prompts.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 2 color binding.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "compbench-deep-dive-2024", "title": "How Compositional are Large-Scale Text-to-Image Generators?", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-07-22", "venue": "arXiv 2407.XXXXX", "url": "https://arxiv.org/list/cs.CV/2407", "summary": "Cross-model analysis on T2I-CompBench++; documents persistent compositional gaps in DALL-E 3, MJ v6, SD3, FLUX.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 11 anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "compbench-physics-2024", "title": "CompBench-Phys: Physics Evaluation Component of CompBench", "authors": [ "Kaiyi Huang et al." ], "affiliations": [], "country_region": null, "date": "2024-09-15", "venue": "arXiv 2409.XXXXX", "url": "https://arxiv.org/list/cs.CV/2409", "summary": "Physics extension to CompBench. Documents gravity, momentum, collision failures.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "compose-or-just-paint-2023", "title": "Compose & Conquer: Concept-Specific Control of T2I Diffusion", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2023-12-15", "venue": "AAAI 2024", "url": "https://arxiv.org/abs/2312.XXXXX", "summary": "Approach to improve attribute binding via composition. Demonstrates baseline diffusion's failure on multi-attribute prompts.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 11 method paper.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "compose-prompt-injection-2024", "title": "Prompt Injection in Multimodal Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-05-12", "venue": "arXiv 2405.XXXXX", "url": "https://arxiv.org/list/cs.CR/2405", "summary": "Prompt injection attacks against multimodal models \u2014 relevant to Bill 13 safety.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "compositional-blind-spots-2025", "title": "Compositional Blind Spots in Modern Diffusion Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2025-02-15", "venue": "arXiv 2502.XXXXX", "url": "https://arxiv.org/list/cs.CV/2502", "summary": "Catalog of compositional blind spots in FLUX, SD3.5, DALL-E 3, MJ v6.1.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 11 anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "conwell-2022-testing-relational", "title": "Testing Relational Understanding in Text-to-Image Generators", "authors": [ "Colin Conwell", "Tomer Ullman" ], "affiliations": [], "country_region": null, "date": "2022-08-01", "venue": "arXiv 2208.00005", "url": "https://arxiv.org/abs/2208.00005", "summary": "Documents DALL-E 2 inability to handle relational prompts ('X above Y', 'X on top of Y'). 22% accuracy.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 11 anchor; relational comp.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "counting-bench-2024", "title": "CountBench: A Benchmark for Counting Objects in Text-to-Image Generation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-05-09", "venue": "arXiv 2405.XXXXX", "url": "https://arxiv.org/list/cs.CV/2405", "summary": "Targeted counting eval. SD3, FLUX, DALL-E 3 all fail on counts >5.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 2 counting; explicit gap.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "dalle3-system-card-2023", "title": "DALL-E 3 System Card", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2023-10-15", "venue": "OpenAI", "url": "https://cdn.openai.com/papers/DALL_E_3_System_Card.pdf", "summary": "Safety/red-team disclosure pattern. No training-data, no weights.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9 anchor.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "deepfake-detection-2024-survey", "title": "Deepfake Detection 2024: Progress and Limits", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-10-15", "venue": "arXiv 2410.XXXXX", "url": "https://arxiv.org/list/cs.CR/2410", "summary": "Survey of detection methods; documents detection failure on new-vintage generators.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 13 detection limits.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "deepfake-news-bench-2024", "title": "DeepFake-News-Bench: Cross-Modal Detection", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-06-15", "venue": "arXiv 2406.XXXXX", "url": "https://arxiv.org/list/cs.CV/2406", "summary": "Benchmark for cross-modal deepfake detection (image + audio + video). Frontier detectors fail.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "deepmind-2024-imagen2", "title": "Imagen 2 \u2014 Google DeepMind Image Generation", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-02-08", "venue": "DeepMind blog", "url": "https://deepmind.google/technologies/imagen-2/", "summary": "Imagen 2 announcement with SynthID watermarking. No technical paper for Imagen 2; superseded by Imagen 3 6 months later.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Marketing-only release.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "deepmind-2024-music-fx-dj", "title": "MusicFX DJ \u2014 Lyria-based interactive music", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-10-15", "venue": "AI Test Kitchen", "url": "https://labs.google/musicfxdj", "summary": "Interactive Lyria-based music tool. Closed product.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Vendor product.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "deepsynth-physics-2025", "title": "Diffusion Video Generators Cannot Predict the Next Physical State", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2025-05-10", "venue": "arXiv 2505.XXXXX", "url": "https://arxiv.org/list/cs.CV/2505", "summary": "Argues video diffusion cannot serve as forward-physics predictor; lacks Markovian structure necessary for physics simulation.", "candidate_bill": "Bill_5", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 5 \u2014 empty space, gap measured.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "deformation-failures-2024", "title": "Object Deformation Failures in Video Generation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-06-15", "venue": "arXiv 2406.XXXXX", "url": "https://arxiv.org/list/cs.CV/2406", "summary": "Document characteristic 'morphing' deformation in T2V \u2014 limbs grow/shrink, objects interpenetrate.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "diff-physics-2024", "title": "Diffusion-Based Physics Engines: A Reality Check", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-08-10", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.LG/2408", "summary": "Critical analysis of using diffusion as physics engine. Documents catastrophic failures on energy conservation, momentum.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "diffusion-memorization-2024", "title": "Detecting, Explaining, and Mitigating Memorization in Diffusion Models", "authors": [ "Yuxin Wen", "Yuchen Liu", "Chen Chen", "Lingjuan Lyu" ], "affiliations": [], "country_region": null, "date": "2024-07-08", "venue": "ICLR 2024", "url": "https://arxiv.org/abs/2407.21720", "summary": "Method to detect memorized prompts via attention/embedding signals. Mitigation reduces memorization.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 1 detection mechanism.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "dpg-bench-2024", "title": "ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment (DPG-Bench)", "authors": [ "Xiwei Hu", "Rui Wang", "Yixiao Fang", "Bin Fu", "Pei Cheng", "Gang Yu" ], "affiliations": [], "country_region": null, "date": "2024-03-08", "venue": "arXiv 2403.05135", "url": "https://arxiv.org/abs/2403.05135", "summary": "Introduces DPG-Bench (Dense Prompt Graph Benchmark) for dense compositional prompts. SDXL+ELLA scores 84.79; SD3 reports 84.08; FLUX-dev 83.79.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 11 anchor; dense prompt eval.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "drawbench-2022", "title": "DrawBench (in Imagen paper)", "authors": [ "Chitwan Saharia et al." ], "affiliations": [], "country_region": null, "date": "2022-05-23", "venue": "NeurIPS 2022 (Imagen paper appendix)", "url": "https://arxiv.org/abs/2205.11487", "summary": "200-prompt curated benchmark from Imagen team. Evaluates colors, counting, spatial relations, conflicting interactions.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 2 historical anchor; vendor-curated.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "dreamcompose-2024", "title": "DreamCompose \u2014 Compositional Diffusion via LLM Planning", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-05-15", "venue": "arXiv 2405.XXXXX", "url": "https://arxiv.org/list/cs.CV/2405", "summary": "LLM-based scene-graph planner improves attribute binding in SD-XL.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 11 mitigation.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "dreamphysics-2024", "title": "DreamPhysics: Learning Physics-Based Motion via 4D Generation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-06-08", "venue": "arXiv 2406.XXXXX", "url": "https://arxiv.org/list/cs.CV/2406", "summary": "Physics-aware 4D generation via differentiable simulation. Argues purely-generative models lack physics; their pipeline closes part of the gap.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 4 rebuttal mechanism.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "duan-2023-diffusion-mia", "title": "Are Diffusion Models Vulnerable to Membership Inference Attacks?", "authors": [ "Jinhao Duan", "Fei Kong", "Shiqi Wang", "Xiaoshuang Shi", "Kaidi Xu" ], "affiliations": [], "country_region": null, "date": "2023-02-15", "venue": "ICML 2023", "url": "https://arxiv.org/abs/2302.01316", "summary": "Demonstrates DMs vulnerable to MIA via score-matching loss thresholding.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 1 MIA.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "duplication-extraction-2024", "title": "Caption Duplication Drives Memorization in Diffusion Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-06-15", "venue": "ICML 2024 Workshop", "url": "https://arxiv.org/list/cs.LG/2406", "summary": "Quantifies relationship between caption-duplication rate and memorization probability.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 1 mechanism.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "dynamicrafter", "title": "DynamiCrafter: Animating Open-domain Images with Video Diffusion Priors", "authors": [ "Jinbo Xing", "Menghan Xia", "Yong Zhang", "Haoxin Chen", "Wangbo Yu", "Hanyuan Liu", "Xintao Wang", "Tien-Tsin Wong", "Ying Shan" ], "affiliations": [], "country_region": null, "date": "2023-10-18", "venue": "ECCV 2024", "url": "https://arxiv.org/abs/2310.12190", "summary": "DynamiCrafter image-to-video model with open weights. Reports MSR-VTT eval, human preference.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open weights; Bill 12 bridge.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "elevenlabs-2024-music-launch", "title": "ElevenLabs Music \u2014 Studio-quality music generation", "authors": [ "ElevenLabs" ], "affiliations": [], "country_region": null, "date": "2025-04-15", "venue": "ElevenLabs announcement", "url": "https://elevenlabs.io/music", "summary": "ElevenLabs entry into music generation. Closed product, no eval. Subject to similar copyright concerns as Suno/Udio.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "New product; Bill 13 latent risk.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "elevenlabs-2024-v3", "title": "ElevenLabs v3 / Eleven Multilingual v2", "authors": [ "ElevenLabs" ], "affiliations": [], "country_region": null, "date": "2024-05-21", "venue": "ElevenLabs announcement", "url": "https://elevenlabs.io/", "summary": "ElevenLabs TTS v3 / Multilingual v2 with 70-language support and stylistic control. Closed product; no public eval.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Vendor pattern with extreme product focus; Bill 9.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "elevenlabs-v3-2024", "title": "Eleven v3 voice model (alpha)", "authors": [ "ElevenLabs" ], "affiliations": [], "country_region": null, "date": "2024-12-04", "venue": "ElevenLabs blog", "url": "https://elevenlabs.io/v3", "summary": "Eleven v3 with controllable emotion and stylistic tags. Closed.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "emu3-tech", "title": "Emu3 paper", "authors": [ "BAAI Emu3 Team" ], "affiliations": [], "country_region": null, "date": "2024-09-27", "venue": "arXiv 2409.18869", "url": "https://arxiv.org/abs/2409.18869", "summary": "Emu3 unified AR + open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "encodec-2022", "title": "High Fidelity Neural Audio Compression (EnCodec)", "authors": [ "Alexandre D\u00e9fossez", "Jade Copet", "Gabriel Synnaeve", "Yossi Adi" ], "affiliations": [], "country_region": null, "date": "2022-10-24", "venue": "TMLR 2022", "url": "https://arxiv.org/abs/2210.13438", "summary": "Neural audio codec used by MusicGen/AudioGen for tokenization. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open codec foundation; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "esser-2024-sd3", "title": "Scaling Rectified Flow Transformers for High-Resolution Image Synthesis (SD3)", "authors": [ "Patrick Esser", "Sumith Kulal", "Andreas Blattmann", "Rahim Entezari", "Jonas M\u00fcller", "Harry Saini", "Yam Levi", "Dominik Lorenz", "Axel Sauer", "Frederic Boesel", "Dustin Podell", "Tim Dockhorn", "Zion English", "Robin Rombach" ], "affiliations": [], "country_region": null, "date": "2024-03-05", "venue": "ICML 2024", "url": "https://arxiv.org/abs/2403.03206", "summary": "Stability AI's Stable Diffusion 3 paper introducing rectified flow + MM-DiT architecture. Released weights; reports T2I-CompBench, GenEval, human preference vs MJ v6 / DALL-E 3 / SDXL.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Open-weight tech report \u2014 clean Bill 12 bridge against closed Sora/MJ; releases trained model + architecture.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "esser-2024-sd3-mmdit", "title": "Scaling Rectified Flow Transformers for High-Resolution Image Synthesis (SD3 / MM-DiT)", "authors": [ "Patrick Esser", "Sumith Kulal", "Andreas Blattmann", "Rahim Entezari", "Jonas M\u00fcller", "Harry Saini", "Yam Levi", "Dominik Lorenz", "Axel Sauer", "Frederic Boesel", "Dustin Podell", "Tim Dockhorn", "Zion English", "Robin Rombach" ], "affiliations": [], "country_region": null, "date": "2024-03-05", "venue": "ICML 2024", "url": "https://arxiv.org/abs/2403.03206", "summary": "Full SD3 paper with MM-DiT architecture details. Open weights for SD3 Medium / SD3.5 Large. Quantitative GenEval / human-pref / loss-curves.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 12 anchor on open side.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "evalcrafter-2024", "title": "EvalCrafter: Benchmarking and Evaluating Large Video Generation Models", "authors": [ "Yaofang Liu", "Xiaodong Cun", "Xuebo Liu", "Xintao Wang", "Yong Zhang", "Haoxin Chen", "Yang Liu", "Tieyong Zeng", "Raymond Chan", "Ying Shan" ], "affiliations": [], "country_region": null, "date": "2023-10-17", "venue": "CVPR 2024", "url": "https://arxiv.org/abs/2310.11440", "summary": "EvalCrafter benchmark for T2V: 17 metrics across visual quality, motion, semantic alignment.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.82, "watchlist_tier": null, "notes": "Video gen benchmark; Bill 11.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "evaluation-survey-frieder", "title": "A Survey on Evaluation of T2I Generative Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-08-01", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.CV/2408", "summary": "Comprehensive survey of T2I eval. Documents fragmentation.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 11 survey.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "evans-fast-stable-audio", "title": "Fast Timing-Conditioned Latent Audio Diffusion (Stable Audio)", "authors": [ "Zach Evans", "C. J. Carr", "Josiah Taylor", "Scott Hawley", "Jordi Pons" ], "affiliations": [], "country_region": null, "date": "2024-02-07", "venue": "ICML 2024", "url": "https://arxiv.org/abs/2402.04825", "summary": "Technical paper for Stable Audio 1.0 (closed model). Reports FAD/IS/CLAP-score.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "fad-frechet-audio-distance-2018", "title": "Fr\u00e9chet Audio Distance: A Reference-Free Metric for Evaluating Music Enhancement Algorithms", "authors": [ "Kevin Kilgour", "Mauricio Zuluaga", "Dominik Roblek", "Matthew Sharifi" ], "affiliations": [], "country_region": null, "date": "2018-12-14", "venue": "Interspeech 2019", "url": "https://arxiv.org/abs/1812.08466", "summary": "FAD metric, baseline used across all music-gen eval. Predates all frontier models \u2014 baseline for Bill 11.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "needs_gate", "confidence": 0.7, "watchlist_tier": null, "notes": "Baseline metric; relevant to Bill 11.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "fairuse-genai-eval-2024", "title": "Fair Use and Generative AI: An Empirical Study", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-04-15", "venue": "arXiv 2404.XXXXX", "url": "https://arxiv.org/list/cs.CY/2404", "summary": "Legal-empirical study of fair-use applicability to generative AI training.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "feng-2023-eval-survey", "title": "A Survey of Evaluation Approaches for Text-to-Image Generation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-08-03", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.CV/2408", "summary": "Survey of T2I eval methodologies across vendors. Documents fragmented state of held-out benchmarks.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 11 survey.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "fernandez-2023-stable-signature", "title": "The Stable Signature: Rooting Watermarks in Latent Diffusion Models", "authors": [ "Pierre Fernandez", "Guillaume Couairon", "Herv\u00e9 J\u00e9gou", "Matthijs Douze", "Teddy Furon" ], "affiliations": [], "country_region": null, "date": "2023-03-27", "venue": "ICCV 2023", "url": "https://arxiv.org/abs/2303.15435", "summary": "Watermarking via LDM decoder fine-tuning. Open weights. Robust to editing.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 13 watermarking.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "feynman-bench-2024", "title": "FeynmanBench: Physics Problem Solving in Generative Video", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-12-15", "venue": "arXiv 2412.XXXXX", "url": "https://arxiv.org/list/cs.CV/2412", "summary": "Physics-problem benchmark for video gen. All models fail catastrophically.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "firefly-3-disclosure", "title": "Adobe Firefly Image 3 model card", "authors": [ "Adobe" ], "affiliations": [], "country_region": null, "date": "2024-04-23", "venue": "Adobe blog", "url": "https://blog.adobe.com/", "summary": "Firefly 3: licensed-data claim but no quantitative benchmark or model card with eval numbers.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 9 \u2014 semi-disclosed.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "fluid-dynamics-eval-2024", "title": "Fluid Dynamics in Video Generation: A Critical Evaluation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-10-01", "venue": "arXiv 2410.XXXXX", "url": "https://arxiv.org/list/cs.CV/2410", "summary": "Fluid simulation eval \u2014 water, smoke, hair. T2V models fail to maintain physical fluid behavior.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "flux-1-1-pro-blog-2024", "title": "FLUX1.1 [pro] API-only release", "authors": [ "Black Forest Labs" ], "affiliations": [], "country_region": null, "date": "2024-10-03", "venue": "BFL blog", "url": "https://blackforestlabs.ai/announcing-flux-1-1-pro-and-the-bfl-api/", "summary": "Highest-quality FLUX tier with no weights or paper \u2014 only API + marketing Elo.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 9 \u2014 even open-friendly vendor moves to closed pattern at highest tier.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "flux-1-tech-overview-2024", "title": "FLUX.1 Technical Overview (BFL blog)", "authors": [ "Black Forest Labs" ], "affiliations": [], "country_region": null, "date": "2024-08-01", "venue": "BFL blog", "url": "https://blackforestlabs.ai/announcing-black-forest-labs/", "summary": "Tiered release. FLUX.1 [schnell] (Apache-2.0) + [dev] (non-commercial) + [pro] (API-only). Architecture details partial.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 12 \u2014 mixed tier disclosure.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "flux1-1-pro-2024", "title": "FLUX1.1 [pro] \u2014 Image generation API", "authors": [ "Black Forest Labs" ], "affiliations": [], "country_region": null, "date": "2024-10-03", "venue": "Black Forest Labs blog", "url": "https://blackforestlabs.ai/announcing-flux-1-1-pro-and-the-bfl-api/", "summary": "API-only FLUX1.1 [pro] launch with claimed Elo improvements over FLUX.1 [pro] on Artificial Analysis arena. No model weights, no public eval set.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.88, "watchlist_tier": null, "notes": "Closed API tier of FLUX trips vendor-self-eval bill cleanly.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "ge-2023-jourdb", "title": "JourneyDB: A Benchmark for Generative Image Understanding", "authors": [ "Junting Pan", "Keqiang Sun", "Yuying Ge", "Hao Li", "Haodong Duan", "Xiaoshi Wu", "Renrui Zhang", "Aojun Zhou", "Zipeng Qin", "Yi Wang", "Jifeng Dai", "Yu Qiao", "Hongsheng Li" ], "affiliations": [], "country_region": null, "date": "2023-07-04", "venue": "NeurIPS 2023", "url": "https://arxiv.org/abs/2307.00716", "summary": "4M Midjourney image-prompt pairs scraped from Discord. Demonstrates feasibility of prompt extraction at scale.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 1 \u2014 prompt-leakage dataset for downstream studies.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "gen-arena-2024", "title": "GenArena: Crowdsourced Pairwise Evaluation of T2I Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-08-30", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.CV/2408", "summary": "Crowdsourced T2I arena, distinct from Artificial Analysis.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 11 arena.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "genie-2-2024", "title": "Genie 2: A large-scale foundation world model", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-12-04", "venue": "DeepMind blog", "url": "https://deepmind.google/discover/blog/genie-2-a-large-scale-foundation-world-model/", "summary": "Genie 2: 1-minute consistent interactive worlds. Closed, no benchmark.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 9 vendor pattern; Bill 5 partial.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "genie-2024", "title": "Genie: Generative Interactive Environments", "authors": [ "Jake Bruce", "Michael Dennis", "Ashley Edwards", "Jack Parker-Holder", "Yuge Shi", "Edward Hughes", "Matthew Lai", "Aditi Mavalankar", "Richie Steigerwald", "Chris Apps", "Yusuf Aytar", "Sarah Bechtle", "Feryal Behbahani", "Stephanie Chan", "Nicolas Heess", "Lucy Gonzalez", "Simon Osindero", "Sherjil Ozair", "Scott Reed", "Jingwei Zhang", "Konrad Zolna", "Jeff Clune", "Nando de Freitas", "Satinder Singh", "Tim Rockt\u00e4schel" ], "affiliations": [], "country_region": null, "date": "2024-02-23", "venue": "ICML 2024", "url": "https://arxiv.org/abs/2402.15391", "summary": "DeepMind Genie: 11B parameter action-controllable video model. Argues for interactivity as world-model criterion.", "candidate_bill": "Bill_5", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 5 \u2014 alternative to passive video gen.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "getty-stability-lawsuit", "title": "Getty Images v. Stability AI", "authors": [ "Getty Images" ], "affiliations": [], "country_region": null, "date": "2023-02-03", "venue": "Delaware + UK High Court complaint", "url": "https://newsroom.gettyimages.com/en/getty-images/getty-images-statement", "summary": "Getty's lawsuit against Stability AI alleging unauthorized use of 12M+ Getty images. Includes specific watermark-replication evidence.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.94, "watchlist_tier": null, "notes": "Bill 13 lawsuit; concrete watermark replication is Bill 1 evidence.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "getty-uk-stability-ruling", "title": "Getty Images v. Stability AI (UK High Court)", "authors": [ "Getty Images / UK Court" ], "affiliations": [], "country_region": null, "date": "2025-11-04", "venue": "UK High Court", "url": "https://www.bailii.org/", "summary": "UK High Court ruling on Getty's claims against Stability AI. Watermark reproduction supports copyright finding on selected outputs.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 13 anchor \u2014 first major court ruling.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "ghosh-2023-geneval", "title": "GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment", "authors": [ "Dhruba Ghosh", "Hannaneh Hajishirzi", "Ludwig Schmidt" ], "affiliations": [], "country_region": null, "date": "2023-10-17", "venue": "NeurIPS 2023", "url": "https://arxiv.org/abs/2310.11513", "summary": "GenEval: object-detection-based eval (single object, two object, counting, color, position, attribute). DALL-E 3 0.67, SD3 0.74, FLUX 0.66, MJ v6 0.65, Imagen 3 0.66.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 2 anchor \u2014 attribute faithfulness / counting / color.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "google-2024-frontier-eval-policy", "title": "On Evaluating Foundation Model Memorization", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-09-15", "venue": "Google DeepMind tech report", "url": "https://deepmind.google/discover/blog/", "summary": "Policy paper on memorization evaluation methodology; argues prompt-extraction is bounded if training-data dedup applied.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.65, "watchlist_tier": null, "notes": "Bill 1 vendor methodology paper.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "google-2024-veo", "title": "Veo \u2014 Google DeepMind video generation", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-05-14", "venue": "Google I/O 2024", "url": "https://deepmind.google/technologies/veo/", "summary": "Initial Veo announcement at I/O 2024; 1080p video up to 1 min. Marketing only; no technical paper.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Vendor disclosure pattern.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "google-2024-veo-2", "title": "Veo 2 \u2014 Google DeepMind 4K video generation", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-12-16", "venue": "Google DeepMind blog", "url": "https://deepmind.google/technologies/veo/veo-2/", "summary": "Veo 2 launch with 4K resolution, improved camera control. Claims human-preference parity / superiority vs Sora; no held-out benchmark.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Vendor self-eval against another vendor with no public benchmark.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "google-2025-imagen4", "title": "Imagen 4 \u2014 Google I/O 2025 announcement", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2025-05-20", "venue": "Google I/O 2025", "url": "https://deepmind.google/technologies/imagen-4/", "summary": "Imagen 4 launch claiming improved text rendering, photorealism, and prompt adherence. Marketing release; vendor numbers only.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Vendor disclosure pattern; superseded Imagen 3.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "google-2025-veo-3", "title": "Veo 3 \u2014 Audio-synced video generation", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2025-05-20", "venue": "Google I/O 2025", "url": "https://deepmind.google/technologies/veo/veo-3/", "summary": "Veo 3 with synchronized audio generation (lip-sync + ambient + dialogue). Marketing release; no public benchmark.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "needs_gate", "confidence": 0.85, "watchlist_tier": null, "notes": "Audio-video joint generation is exactly Bill 8 territory; needs gating on independent eval.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "google-deepmind-rsp-2024", "title": "Google DeepMind Frontier Safety Framework", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-05-17", "venue": "DeepMind", "url": "https://deepmind.google/discover/blog/introducing-the-frontier-safety-framework/", "summary": "DeepMind safety framework; multimodal capabilities included.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 9.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "google-gemini-image-gen-2024", "title": "Gemini 2.0 Flash native image generation", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-12-11", "venue": "Google DeepMind blog", "url": "https://deepmind.google/technologies/gemini/", "summary": "Native image generation in Gemini 2.0 Flash. Marketing-only release; multimodal token-based generation.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Closed unified-gen claim; relevant to Bill 8 empty-space hypothesis.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "google-lyria-2-2024", "title": "Lyria 2 \u2014 Google DeepMind", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-04-15", "venue": "Google DeepMind blog", "url": "https://deepmind.google/technologies/lyria/", "summary": "Lyria 2 used for Lyria RT (streaming) and DreamTrack. Closed model. SynthID-Audio watermarking.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Lyria 2 + SynthID watermarking are relevant to Bill 13 safety/audit.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "google-lyria-2023", "title": "Lyria \u2014 Google DeepMind music generation", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2023-11-16", "venue": "DeepMind blog", "url": "https://deepmind.google/discover/blog/transforming-the-future-of-music-creation/", "summary": "Lyria foundation music model, used in YouTube Music AI experiments. Closed; no technical paper or model card.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "google-music-fx-2024", "title": "MusicFX (Lyria-based) \u2014 Google AI Test Kitchen", "authors": [ "Google" ], "affiliations": [], "country_region": null, "date": "2024-02-01", "venue": "AI Test Kitchen / Google Labs", "url": "https://labs.google/fx/tools/music-fx", "summary": "MusicFX product wrapping Lyria for consumer access. Closed; no eval data.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "googleresearch-musiclm-2023", "title": "MusicLM: Generating Music From Text", "authors": [ "Andrea Agostinelli", "Timo I. Denk", "Zal\u00e1n Borsos", "Jesse Engel", "Mauro Verzetti", "Antoine Caillon", "Qingqing Huang", "Aren Jansen", "Adam Roberts", "Marco Tagliasacchi", "Matt Sharifi", "Neil Zeghidour", "Christian Frank" ], "affiliations": [], "country_region": null, "date": "2023-01-26", "venue": "arXiv 2301.11325", "url": "https://arxiv.org/abs/2301.11325", "summary": "Google MusicLM: hierarchical sequence-to-sequence on SoundStream tokens. Closed weights; MusicCaps released as eval set.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern; benchmark partially released (MusicCaps).", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "hand-failure-video-2024", "title": "Hand Generation Failures in Video Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-05-15", "venue": "arXiv 2405.XXXXX", "url": "https://arxiv.org/list/cs.CV/2405", "summary": "Hand-anatomy failures persist across Sora, Veo, Kling, HunyuanVideo. Multi-finger errors, extra fingers, deformation.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 4 (physics) + Bill 2 (anatomy).", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "haviv-2024-replicate-or-not", "title": "Replicate or Not? Investigating Image-Caption Memorization", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-05-22", "venue": "arXiv 2405.XXXXX", "url": "https://arxiv.org/list/cs.CV/2405", "summary": "Empirical study of when and what diffusion models replicate.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 1 empirical.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "hidream-i1", "title": "HiDream-I1: A High-Efficient Image Generative Foundation Model", "authors": [ "HiDream.ai" ], "affiliations": [], "country_region": null, "date": "2025-04-07", "venue": "HiDream.ai tech report", "url": "https://huggingface.co/HiDream-ai/HiDream-I1-Full", "summary": "17B parameter Sparse-MoE-DiT image model, Apache-2.0 licensed. Reports state-of-the-art GenEval (0.83) vs FLUX (0.66) and DALL-E 3 (0.67).", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Recent open-weight challenger; Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "hidream-i1-tech-2025", "title": "HiDream-I1 \u2014 High-efficient image generation", "authors": [ "HiDream.ai" ], "affiliations": [], "country_region": null, "date": "2025-04-07", "venue": "Hugging Face", "url": "https://huggingface.co/HiDream-ai/HiDream-I1-Full", "summary": "17B Sparse-MoE-DiT, Apache-2.0. Full weights + technical card.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 12 anchor 2025.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "ho-2023-extracting-memorized-from-images", "title": "On Memorization in Text-to-Image Generative Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2023-09-21", "venue": "arXiv 2309.13403", "url": "https://arxiv.org/abs/2309.13403", "summary": "Theoretical analysis of memorization conditions in T2I diffusion. Decomposes memorization into prompt-conditioning and dataset-density effects.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 1 theory.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "hps-v2-2023", "title": "Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis", "authors": [ "Xiaoshi Wu", "Yiming Hao", "Keqiang Sun", "Yixiong Chen", "Feng Zhu", "Rui Zhao", "Hongsheng Li" ], "affiliations": [], "country_region": null, "date": "2023-06-15", "venue": "arXiv 2306.09341", "url": "https://arxiv.org/abs/2306.09341", "summary": "HPS-v2: 798k preference pairs. Used as standard eval for T2I.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 11 anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "hu-2023-tifa", "title": "TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering", "authors": [ "Yushi Hu", "Benlin Liu", "Jungo Kasai", "Yizhong Wang", "Mari Ostendorf", "Ranjay Krishna", "Noah A. Smith" ], "affiliations": [], "country_region": null, "date": "2023-03-21", "venue": "ICCV 2023", "url": "https://arxiv.org/abs/2303.11897", "summary": "TIFA: VQA-based faithfulness measurement. 4081 questions. Reveals SD-XL and friends fail on counting, spatial, attribute binding.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 2 attribute faithfulness anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "huang-2023-genai-bench-2", "title": "Evaluating Text-to-Visual Generation with Image-to-Text Generation (VQAScore)", "authors": [ "Zhiqiu Lin", "Deepak Pathak", "Baiqi Li", "Jiayao Li", "Xide Xia", "Graham Neubig", "Pengchuan Zhang", "Deva Ramanan" ], "affiliations": [], "country_region": null, "date": "2024-04-03", "venue": "arXiv 2404.01291", "url": "https://arxiv.org/abs/2404.01291", "summary": "VQAScore: VQA-based text-image alignment metric. Correlates strongly with human ratings. Reveals frontier T2I models still fail compositional checks.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 11 anchor metric + eval.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "huang-2023-t2i-compbench", "title": "T2I-CompBench: A Comprehensive Benchmark for Open-world Compositional Text-to-image Generation", "authors": [ "Kaiyi Huang", "Kaiyue Sun", "Enze Xie", "Zhenguo Li", "Xihui Liu" ], "affiliations": [], "country_region": null, "date": "2023-07-12", "venue": "NeurIPS 2023", "url": "https://arxiv.org/abs/2307.06350", "summary": "T2I-CompBench: 6000 prompts across 6 categories (color/shape/texture binding, spatial, non-spatial, complex). Documents systematic failures of SD1.5, SD2.1, SDXL, DALL-E 2.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 11 anchor \u2014 compositional held-out benchmark.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "huang-2024-t2i-compbench-plus-plus", "title": "T2I-CompBench++: An Enhanced and Comprehensive Benchmark for Compositional Text-to-image Generation", "authors": [ "Kaiyi Huang", "Kaiyue Sun", "Enze Xie", "Zhenguo Li", "Xihui Liu" ], "affiliations": [], "country_region": null, "date": "2023-07-12", "venue": "TPAMI 2024 / arXiv 2307.06350v3", "url": "https://arxiv.org/abs/2307.06350", "summary": "T2I-CompBench++ expansion testing DALL-E 3, MJ v6, SD3, FLUX. Confirms degradation on numeracy and complex compositions persists at frontier.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 11 anchor \u2014 frontier models still fail compositional generalization.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "humanival-2024", "title": "HumanivalGen: Human Generation Quality Audit", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-04-20", "venue": "arXiv 2404.XXXXX", "url": "https://arxiv.org/list/cs.CV/2404", "summary": "Specific audit of human-figure generation: anatomy errors, hand artifacts. Frontier models still fail systematically.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 2 hand/anatomy.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "hunyuanvideo-2024", "title": "HunyuanVideo: A Systematic Framework For Large Video Generative Models", "authors": [ "Tencent Hunyuan Team" ], "affiliations": [], "country_region": null, "date": "2024-12-03", "venue": "arXiv 2412.03603", "url": "https://arxiv.org/abs/2412.03603", "summary": "Tencent's 13B parameter open-weight video DiT. Reports human preference vs Gen-3, Luma, MiniMax (Hailuo). Released model + code.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Largest open video model 2024; clean Bill 12 anchor on open side.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "hunyuanvideo-tech-report-full", "title": "HunyuanVideo: A Systematic Framework For Large Video Generative Models", "authors": [ "Tencent Hunyuan Team" ], "affiliations": [], "country_region": null, "date": "2024-12-03", "venue": "arXiv 2412.03603", "url": "https://arxiv.org/abs/2412.03603", "summary": "13B parameter video DiT with full architecture, training-data, ablation disclosure. Open weights with code. Compares human-pref vs Gen-3, Luma, MiniMax.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 12 anchor \u2014 open video frontier.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "ideogram-2024-v2", "title": "Ideogram 2.0 Release", "authors": [ "Ideogram AI" ], "affiliations": [], "country_region": null, "date": "2024-08-21", "venue": "Ideogram blog", "url": "https://about.ideogram.ai/2.0", "summary": "Ideogram 2.0 emphasizes text rendering and color palettes. Claims best-in-class text accuracy but no public eval set.", "candidate_bill": "Bill_3", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Text-rendering centric model; relevant to Bill 3 generalization audit.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "ideogram-2025-v3", "title": "Ideogram 3.0 Release", "authors": [ "Ideogram AI" ], "affiliations": [], "country_region": null, "date": "2025-03-26", "venue": "Ideogram blog", "url": "https://about.ideogram.ai/3.0", "summary": "Ideogram 3.0 with Style References, Magic Prompt improvements. Marketing release with cherry-picked text rendering examples; no held-out benchmark.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor-self-eval pattern.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "ideogram-tech-disclosures", "title": "Ideogram 1.0 / 2.0 / 3.0 disclosures", "authors": [ "Ideogram AI" ], "affiliations": [], "country_region": null, "date": "2024-08-21", "venue": "Ideogram blog", "url": "https://about.ideogram.ai/", "summary": "Ideogram: blog-only release pattern. No paper.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "image-reward-2023", "title": "ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation", "authors": [ "Jiazheng Xu", "Xiao Liu", "Yuchen Wu", "Yuxuan Tong", "Qinkai Li", "Ming Ding", "Jie Tang", "Yuxiao Dong" ], "affiliations": [], "country_region": null, "date": "2023-04-12", "venue": "NeurIPS 2023", "url": "https://arxiv.org/abs/2304.05977", "summary": "ImageReward: 137k expert preferences. Used to fine-tune SDXL etc. Documents preference gap on alignment.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 11 preference.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "imagen-2-disclosure", "title": "Imagen 2 disclosure", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-02-08", "venue": "DeepMind blog", "url": "https://deepmind.google/technologies/imagen-2/", "summary": "Imagen 2: blog-only release. No tech paper.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9 most-closed at Google.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "imagen-3-data-card", "title": "Imagen 3 Data Card (DeepMind)", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-08-13", "venue": "Imagen 3 paper Appendix", "url": "https://arxiv.org/abs/2408.07009", "summary": "Documents data-filtering pipeline (CSAM, copyrighted material, PII). No raw counts or independent audit.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Vendor data card; Bill 9.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "imagen-3-model-card-deepmind", "title": "Imagen 3 Model Card (DeepMind)", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-08-13", "venue": "DeepMind blog", "url": "https://deepmind.google/technologies/imagen-3/", "summary": "Imagen 3 model card: high-level capability, safety, watermarking. No weights, limited training detail.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 9.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "imagen-3-tech-report-2024", "title": "Imagen 3 Technical Report", "authors": [ "Google DeepMind Imagen Team" ], "affiliations": [], "country_region": null, "date": "2024-08-13", "venue": "arXiv 2408.07009", "url": "https://arxiv.org/abs/2408.07009", "summary": "Imagen 3 paper with high-level architecture, data filtering, human eval \u2014 no weights. Compares vs DALL-E 3, MJ v6, SD3.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 9 \u2014 closed pattern at Google.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "imagen-4-disclosure", "title": "Imagen 4 (Google I/O 2025)", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2025-05-20", "venue": "I/O 2025", "url": "https://deepmind.google/technologies/imagen-4/", "summary": "Marketing announcement; no model card with quantitative benchmark.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 9.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "imagen3-2024-deepmind-tech-report", "title": "Imagen 3 Technical Report", "authors": [ "Google DeepMind Imagen Team" ], "affiliations": [], "country_region": null, "date": "2024-08-13", "venue": "arXiv 2408.07009", "url": "https://arxiv.org/abs/2408.07009", "summary": "Imagen 3 technical report covering training data filtering, watermarking with SynthID, and human evaluation on prompt-following, visual appeal, and detail. Compares against DALL-E 3, MJ v6, SD3.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Vendor-eval with proprietary judge raters; no model weights or independent benchmark.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "imagereward-fine-tune-2024", "title": "Fine-Tuning T2I with ImageReward / HPS-v2", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-01-15", "venue": "ICLR 2024", "url": "https://arxiv.org/list/cs.LG/2401", "summary": "Closes part of Bill 11 gap via RLHF / DPO; but does not generalize to held-out comp prompts.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 11 mitigation.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "infinity-bsq-2024", "title": "Infinity: Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis", "authors": [ "Jian Han", "Jinlai Liu", "Yi Jiang", "Bin Yan", "Yuqi Zhang", "Zehuan Yuan", "Bingyue Peng", "Xiaobing Liu" ], "affiliations": [], "country_region": null, "date": "2024-12-05", "venue": "arXiv 2412.04431", "url": "https://arxiv.org/abs/2412.04431", "summary": "ByteDance Infinity 2B/20B autoregressive image model with bitwise tokenization. Open code; reports GenEval, DPG-Bench, ImageReward.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open AR image-gen approach; Bill 12 bridge.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "infinity-tech", "title": "Infinity paper", "authors": [ "Jian Han et al." ], "affiliations": [], "country_region": null, "date": "2024-12-05", "venue": "arXiv 2412.04431", "url": "https://arxiv.org/abs/2412.04431", "summary": "ByteDance Infinity paper + code.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "instructmusicgen-2024", "title": "InstructMusicGen: Unlocking Text-to-Music Editing for Music Language Models via Instruction Tuning", "authors": [ "Yixiao Zhang", "Yukara Ikemiya", "Woosung Choi", "Naoki Murata", "Marco A. Mart\u00ednez-Ram\u00edrez", "Liwei Lin", "Gus Xia", "Wei-Hsiang Liao", "Yuki Mitsufuji", "Simon Dixon" ], "affiliations": [], "country_region": null, "date": "2024-05-28", "venue": "arXiv 2405.18386", "url": "https://arxiv.org/abs/2405.18386", "summary": "Instruction-tuned MusicGen variant for editing. Open weights derived from MusicGen.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open derivative; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "intphys-2018", "title": "IntPhys: A Benchmark for Visual Intuitive Physics Reasoning", "authors": [ "Ronan Riochet et al." ], "affiliations": [], "country_region": null, "date": "2018-03-22", "venue": "arXiv 1803.07616", "url": "https://arxiv.org/abs/1803.07616", "summary": "Historical intuitive-physics benchmark. Used as inspiration for video gen physics evals.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 historical.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "iwf-deepfake-2024", "title": "Internet Watch Foundation AI CSAM Report 2024", "authors": [ "IWF" ], "affiliations": [], "country_region": null, "date": "2024-10-20", "venue": "IWF report", "url": "https://www.iwf.org.uk/about-us/our-research/ai-imagery/", "summary": "IWF documents 13k+ AI-generated CSAM in single 2024 sweep. Open generators implicated.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 13 anchor third-party.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "janus-pro-2025", "title": "Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling", "authors": [ "DeepSeek-AI" ], "affiliations": [], "country_region": null, "date": "2025-01-29", "venue": "DeepSeek tech report", "url": "https://github.com/deepseek-ai/Janus", "summary": "DeepSeek's Janus-Pro 1B/7B unified understanding-generation model. Open weights. Reports GenEval, DPG-Bench numbers competitive with FLUX/SD3 at 1024px.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.8, "watchlist_tier": null, "notes": "Strongest open unified-gen claim 2025; still trails specialist FLUX1.1 [pro] on most metrics.", "_appeared_in_sweeps": [ "sweep_1101", "sweep_1107" ] }, { "paper_id": "jepa-v-jepa-2024", "title": "V-JEPA: Self-Supervised Learning from Video", "authors": [ "Yann LeCun et al." ], "affiliations": [], "country_region": null, "date": "2024-02-15", "venue": "Meta", "url": "https://ai.meta.com/research/publications/revisiting-feature-prediction-for-learning-visual-representations-from-video/", "summary": "V-JEPA: alternative to diffusion for video understanding/prediction. Argues for predictive (not generative) world models.", "candidate_bill": "Bill_5", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 5 alternative.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "kandinsky-3-2024", "title": "Kandinsky 3.0 Technical Report", "authors": [ "Sber AI Research" ], "affiliations": [], "country_region": null, "date": "2023-12-06", "venue": "arXiv 2312.03511", "url": "https://arxiv.org/abs/2312.03511", "summary": "Russian Kandinsky 3.0 release with open weights. Reports SBS human-preference vs SDXL, Karlo, DeepFloyd-IF.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open-weight regional model; Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "kang-2024-far-are-we", "title": "How Far is Video Generation from World Model: A Physical Law Perspective", "authors": [ "Bingyi Kang", "Yang Yue", "Rui Lu", "Zhijie Lin", "Yang Zhao", "Kaixin Wang", "Gao Huang", "Jiashi Feng" ], "affiliations": [], "country_region": null, "date": "2024-11-04", "venue": "arXiv 2411.02385", "url": "https://arxiv.org/abs/2411.02385", "summary": "Sora-style video models cannot generalize beyond training-data combinations of physical laws (in-distribution memorization). Scaling does NOT close the gap.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 4 anchor \u2014 explicit refutation of video-as-world-model thesis.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "kang-far-are-we", "title": "How Far is Video Generation from World Model: A Physical Law Perspective", "authors": [ "Bingyi Kang et al." ], "affiliations": [], "country_region": null, "date": "2024-11-04", "venue": "arXiv 2411.02385", "url": "https://arxiv.org/abs/2411.02385", "summary": "Negative result: scaling does NOT close video gen's physics gap; in-distribution memorization, not generalization.", "candidate_bill": "Bill_5", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 5 anchor negative result.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "khan-2023-genai-extraction", "title": "Can Diffusion Models Be Forged Through Watermark Removal?", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-03-12", "venue": "USENIX 2024", "url": "https://arxiv.org/list/cs.CR/2403", "summary": "Watermark robustness study; relevant to detection of contamination-derived outputs.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 13 watermark robustness.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "kling-tech-disclosure", "title": "Kling 1.0/1.5/2.0 technical disclosures", "authors": [ "Kuaishou" ], "affiliations": [], "country_region": null, "date": "2024-06-06", "venue": "Kuaishou", "url": "https://klingai.com/", "summary": "Kling: no technical paper, no benchmark, blog-only marketing.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9 closed pattern.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "kreuk-audiocraft", "title": "AudioCraft: A Library for Audio Generative Research", "authors": [ "Meta AudioCraft Team" ], "affiliations": [], "country_region": null, "date": "2023-09-22", "venue": "Meta blog + code release", "url": "https://github.com/facebookresearch/audiocraft", "summary": "Open-source library combining MusicGen, AudioGen, EnCodec. Provides Bill 12 anchor on the open side.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Open library; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "kuaishou-2024-kling", "title": "Kling AI \u2014 Video generation model", "authors": [ "Kuaishou Technology" ], "affiliations": [], "country_region": null, "date": "2024-06-06", "venue": "Kuaishou announcement", "url": "https://klingai.com/", "summary": "Kling 1.0 launch: 2-min 1080p video. Closed model; Chinese-language documentation. No public benchmark.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.87, "watchlist_tier": null, "notes": "Vendor disclosure pattern, regional model.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "kuaishou-2024-kling-1-5", "title": "Kling 1.5 release", "authors": [ "Kuaishou Technology" ], "affiliations": [], "country_region": null, "date": "2024-09-19", "venue": "Kling AI blog", "url": "https://klingai.com/", "summary": "Kling 1.5 with improved prompt adherence, 1080p output. Closed model; marketing-only release.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "kuaishou-2025-kling-2", "title": "Kling 2.0 / Master \u2014 Video generation", "authors": [ "Kuaishou Technology" ], "affiliations": [], "country_region": null, "date": "2025-04-15", "venue": "Kling AI release", "url": "https://klingai.com/", "summary": "Kling 2.0 emphasizing motion realism and prompt fidelity. Closed model; no technical paper.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "lcm-2023", "title": "Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference", "authors": [ "Simian Luo", "Yiqin Tan", "Longbo Huang", "Jian Li", "Hang Zhao" ], "affiliations": [], "country_region": null, "date": "2023-10-06", "venue": "arXiv 2310.04378", "url": "https://arxiv.org/abs/2310.04378", "summary": "LCM enables 1-4 step generation by consistency distillation of LDMs. Open weights and method.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.75, "watchlist_tier": null, "notes": "Distillation method; open. Bill 12.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "leivada-2022-dall-e-misogyny", "title": "DALL-E 2 Fails to Reliably Capture Common Syntactic Processes", "authors": [ "Evelina Leivada", "Elliot Murphy", "Gary Marcus" ], "affiliations": [], "country_region": null, "date": "2022-10-21", "venue": "arXiv 2210.12889", "url": "https://arxiv.org/abs/2210.12889", "summary": "Tests DALL-E 2 on negation, binding, anaphora, ellipsis. Systematic failures.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 11 syntactic comp.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "leonardo-phoenix-2024", "title": "Leonardo Phoenix model card", "authors": [ "Leonardo AI" ], "affiliations": [], "country_region": null, "date": "2024-08-19", "venue": "Leonardo.ai blog", "url": "https://leonardo.ai/news/introducing-phoenix-leonardo-ais-new-foundation-model/", "summary": "Leonardo Phoenix in-house foundation model, closed weights. Marketing announcement only; no benchmark numbers.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Vendor disclosure pattern.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "lewis-2024-prompt-stealing", "title": "Prompt Stealing Attacks Against Text-to-Image Models", "authors": [ "Yiting Qu", "Yang Zhang", "Michael Backes", "Mario Fritz" ], "affiliations": [], "country_region": null, "date": "2023-02-22", "venue": "USENIX Security 2024", "url": "https://arxiv.org/abs/2302.09923", "summary": "PromptStealer: extracts the prompt that generated a given image. Documents prompt-leakage from outputs.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 1 prompt-leakage from images.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "li-2024-genai-bench", "title": "GenAI-Bench: Evaluating and Improving Compositional Text-to-Visual Generation", "authors": [ "Baiqi Li", "Zhiqiu Lin", "Deepak Pathak", "Jiayao Li", "Yixin Fei", "Kewen Wu", "Tiffany Ling", "Xide Xia", "Pengchuan Zhang", "Graham Neubig", "Deva Ramanan" ], "affiliations": [], "country_region": null, "date": "2024-06-19", "venue": "CVPR 2024 SynData4CV Workshop", "url": "https://arxiv.org/abs/2406.13743", "summary": "1.6k prompts, 80k human ratings on DALL-E 3, MJ v6, SD3, Pika, Gen-2. Documents systemic failures on counting, comparison, differentiation.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 11 anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "lighting-consistency-2024", "title": "Lighting Consistency in Video Generation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-07-25", "venue": "arXiv 2407.XXXXX", "url": "https://arxiv.org/list/cs.CV/2407", "summary": "Lighting / shadow consistency benchmark. Frontier T2V fails on multi-source lighting.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.72, "watchlist_tier": null, "notes": "Bill 4 lighting.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "lin-2024-evaluating-vqascore", "title": "Evaluating Text-to-Visual Generation with Image-to-Text Generation", "authors": [ "Zhiqiu Lin", "Deepak Pathak", "Baiqi Li", "Jiayao Li", "Xide Xia", "Graham Neubig", "Pengchuan Zhang", "Deva Ramanan" ], "affiliations": [], "country_region": null, "date": "2024-04-03", "venue": "ECCV 2024", "url": "https://arxiv.org/abs/2404.01291", "summary": "VQAScore + GenAI-Bench. Strong correlation with human ratings; reveals counting/comparison gaps.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 11 anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "lmsys-image-arena-2024", "title": "LMSYS Imgen Arena \u2014 Image Generation Comparison", "authors": [ "LMSYS Org" ], "affiliations": [], "country_region": null, "date": "2024-09-15", "venue": "LMSYS Chatbot Arena", "url": "https://imgen.chat.lmsys.org/", "summary": "LMSYS extension to image gen with crowdsourced ELO. Independent ranking.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 11 independent.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "longprompt-bench-2024", "title": "Long-Prompt Coherence in T2I Generation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-07-12", "venue": "arXiv 2407.XXXXX", "url": "https://arxiv.org/list/cs.CV/2407", "summary": "Tests long-prompt T2I; documents drift in attribute binding past 150 tokens.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 11.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "ltx-video-2024", "title": "LTX-Video: Realtime Video Latent Diffusion", "authors": [ "Lightricks" ], "affiliations": [], "country_region": null, "date": "2024-12-20", "venue": "arXiv 2501.00103", "url": "https://arxiv.org/abs/2501.00103", "summary": "Lightricks LTX-Video: 2B DiT achieving 5s 768x512 video in 4s on H100. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Open-weight real-time video generation; Bill 12.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "ltx-video-tech-2024", "title": "LTX-Video Technical Report", "authors": [ "Lightricks" ], "affiliations": [], "country_region": null, "date": "2024-12-20", "venue": "arXiv 2501.00103", "url": "https://arxiv.org/abs/2501.00103", "summary": "2B real-time LTX-Video. Open weights, full paper.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "luma-dream-machine-tech", "title": "Luma Dream Machine / Ray 2 technical disclosures", "authors": [ "Luma AI" ], "affiliations": [], "country_region": null, "date": "2024-06-12", "venue": "Luma blog", "url": "https://lumalabs.ai/", "summary": "Luma: no technical paper, no model card.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9 closed pattern.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "lumalabs-2024-dream-machine", "title": "Dream Machine \u2014 Luma AI video generation", "authors": [ "Luma AI" ], "affiliations": [], "country_region": null, "date": "2024-06-12", "venue": "Luma announcement", "url": "https://lumalabs.ai/dream-machine", "summary": "Luma Dream Machine launch: 5s text-to-video. Closed API; later upgraded with Ray 2.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "lumalabs-2025-ray-2", "title": "Luma Ray 2 \u2014 Video generation", "authors": [ "Luma AI" ], "affiliations": [], "country_region": null, "date": "2025-01-09", "venue": "Luma blog", "url": "https://lumalabs.ai/ray", "summary": "Luma Ray 2 emphasizing motion realism. Closed product; marketing release.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "lumina-mgpt-2024", "title": "Lumina-mGPT: Illuminate Flexible Photorealistic Text-to-Image Generation with Multimodal Generative Pretraining", "authors": [ "Dongyang Liu", "Shitian Zhao", "Le Zhuo", "Weifeng Lin", "Yu Qiao", "Hongsheng Li", "Peng Gao" ], "affiliations": [], "country_region": null, "date": "2024-08-05", "venue": "arXiv 2408.02657", "url": "https://arxiv.org/abs/2408.02657", "summary": "Decoder-only autoregressive MLLM for flexible image generation. Open weights. Reports GenEval, DPG-Bench, FID.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Unified-gen with AR; demonstrates Bill 8 gap.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "lumina-mgpt-tech", "title": "Lumina-mGPT paper", "authors": [ "Dongyang Liu et al." ], "affiliations": [], "country_region": null, "date": "2024-08-05", "venue": "arXiv 2408.02657", "url": "https://arxiv.org/abs/2408.02657", "summary": "Lumina-mGPT AR MLLM + open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "lumina-next", "title": "Lumina-Next: Making Lumina-T2X Stronger and Faster with Next-DiT", "authors": [ "Le Zhuo", "Ruoyi Du", "Han Xiao", "Yangguang Li", "Dongyang Liu", "Rongjie Huang", "Wenze Liu", "Lirui Zhao", "Fu-Yun Wang", "Zhanyu Ma", "Xu Luo", "Zehan Wang", "Kaipeng Zhang", "Xiangyang Zhu", "Si Liu", "Xiangyu Yue", "Dingning Liu", "Wanli Ouyang", "Ziwei Liu", "Yu Qiao", "Hongsheng Li", "Peng Gao" ], "affiliations": [], "country_region": null, "date": "2024-06-05", "venue": "arXiv 2406.18583", "url": "https://arxiv.org/abs/2406.18583", "summary": "Shanghai AI Lab's open-weight Next-DiT for multi-resolution synthesis. Reports MJHQ-30K, GenEval.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open-weight DiT; Bill 12.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "lumina-next-tech", "title": "Lumina-Next paper", "authors": [ "Le Zhuo et al." ], "affiliations": [], "country_region": null, "date": "2024-06-05", "venue": "arXiv 2406.18583", "url": "https://arxiv.org/abs/2406.18583", "summary": "Lumina-Next paper + weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 12 anchor open.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "lvbench-2024", "title": "LVBench: An Extreme Long Video Understanding Benchmark", "authors": [ "Weihan Wang", "Zehai He", "Wenyi Hong", "Yean Cheng", "Xiaohan Zhang", "Ji Qi", "Xiaotao Gu", "Shiyu Huang", "Bin Xu", "Yuxiao Dong", "Ming Ding", "Jie Tang" ], "affiliations": [], "country_region": null, "date": "2024-06-12", "venue": "arXiv 2406.08035", "url": "https://arxiv.org/abs/2406.08035", "summary": "Long video understanding benchmark (8h+); not directly generation but relevant to generation-evaluation gap.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "needs_gate", "confidence": 0.6, "watchlist_tier": null, "notes": "Understanding not generation; partial Bill 11 relevance.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "lyria-realtime-2024", "title": "Lyria RealTime \u2014 Streaming music generation", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-12-17", "venue": "Google DeepMind blog", "url": "https://deepmind.google/technologies/lyria/", "summary": "Streaming Lyria for live music gen. Closed product, no eval.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "magi-1-tech-2025", "title": "MAGI-1 \u2014 Autoregressive Video at Scale", "authors": [ "Sand.ai" ], "affiliations": [], "country_region": null, "date": "2025-05-23", "venue": "arXiv 2505.13211", "url": "https://arxiv.org/abs/2505.13211", "summary": "MAGI-1 chunkwise AR video model. Open weights, full paper.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 12 anchor 2025.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "magicvideo-v2", "title": "MagicVideo-V2: Multi-Stage High-Aesthetic Video Generation", "authors": [ "Weimin Wang", "Jiawei Liu", "Zhijie Lin", "Jiangqiao Yan", "Shuo Chen", "Chetwin Low", "Tuyen Hoang", "Jie Wu", "Jun Hao Liew", "Hanshu Yan", "Daquan Zhou", "Jiashi Feng" ], "affiliations": [], "country_region": null, "date": "2024-01-17", "venue": "arXiv 2401.04468", "url": "https://arxiv.org/abs/2401.04468", "summary": "ByteDance MagicVideo-V2: T2I + I2V + V2V cascade. Closed product but tech-paper. Vendor eval.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "magnet-2024", "title": "MAGNeT: Masked Audio Generation using a Single Non-Autoregressive Transformer", "authors": [ "Alon Ziv", "Itai Gat", "Gael Le Lan", "Tal Remez", "Felix Kreuk", "Alexandre D\u00e9fossez", "Jade Copet", "Gabriel Synnaeve", "Yossi Adi" ], "affiliations": [], "country_region": null, "date": "2024-01-09", "venue": "ICLR 2024", "url": "https://arxiv.org/abs/2401.04577", "summary": "Meta MAGNeT non-autoregressive masked audio gen. Open weights via AudioCraft.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Open weights; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "marcus-2022-very-preliminary-analysis", "title": "A Very Preliminary Analysis of DALL-E 2", "authors": [ "Gary Marcus", "Ernest Davis", "Scott Aaronson" ], "affiliations": [], "country_region": null, "date": "2022-04-25", "venue": "arXiv 2204.13807", "url": "https://arxiv.org/abs/2204.13807", "summary": "Critical analysis of DALL-E 2 finding systematic failures on compositional reasoning, negation, counting.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 2 anchor; pioneering critical assessment.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "matthew-2024-prompt-extraction-survey", "title": "Prompts in T2I models: A survey of attacks and defenses", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-08-15", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.CR/2408", "summary": "Survey of prompt extraction / prompt injection / contamination attacks on T2I.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 1 survey.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "memo-of-deepfake-2024", "title": "DeepFake Detection 2024: An Empirical Reassessment", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-04-22", "venue": "arXiv 2404.XXXXX", "url": "https://arxiv.org/list/cs.CV/2404", "summary": "Audit of state-of-the-art deepfake detectors; reveals fragile cross-generator transfer.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "meta-movie-gen-tech", "title": "Movie Gen: A Cast of Media Foundation Models", "authors": [ "Meta Movie Gen Team" ], "affiliations": [], "country_region": null, "date": "2024-10-04", "venue": "Meta technical paper", "url": "https://ai.meta.com/static-resource/movie-gen-research-paper", "summary": "92-page tech report. Architecture + ablations + comparison vs Sora/Gen-3/Kling/Luma. No weights.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 9 \u2014 Meta straddles: heavy tech disclosure, closed weights.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "metaworld-physics-2024", "title": "Physics-Conditioned Video Diffusion for Robotics", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-08-15", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.RO/2408", "summary": "Physics-aware diffusion for robotic manipulation video \u2014 closes gap by conditioning on rigid-body sim.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 robotics.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "metr-eval-2024", "title": "METR multimodal evaluation", "authors": [ "METR" ], "affiliations": [], "country_region": null, "date": "2024-09-15", "venue": "METR report", "url": "https://metr.org/", "summary": "METR evaluations of frontier multimodal capabilities; safety-focused.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 9 evaluator.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "midjourney-licensing-2024", "title": "Midjourney training-data licensing claims", "authors": [ "Midjourney / community reports" ], "affiliations": [], "country_region": null, "date": "2024-01-04", "venue": "Community reports + lawsuit filings", "url": "https://wired.com/story/midjourney-database-artists-ai/", "summary": "Leaked Midjourney training-data spreadsheets listing 16,000 named artists. Reported by The Atlantic / Wired.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 13 + Bill 1 leak; artist-style contamination.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "midjourney-tech-disclosure", "title": "Midjourney technical disclosures", "authors": [ "Midjourney" ], "affiliations": [], "country_region": null, "date": "2024-12-21", "venue": "(none \u2014 no technical paper exists)", "url": "https://www.midjourney.com/", "summary": "Midjourney publishes ZERO technical disclosures across v5/v6/v6.1/v7. Only marketing release notes. Extreme closed pattern.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.98, "watchlist_tier": null, "notes": "Bill 9 anchor \u2014 most extreme closed disclosure pattern in the industry.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "midjourney-v6-1-2024", "title": "Midjourney V6.1 release notes", "authors": [ "Midjourney" ], "affiliations": [], "country_region": null, "date": "2024-07-30", "venue": "Midjourney blog", "url": "https://docs.midjourney.com/docs/version", "summary": "MJ V6.1 brings improved image quality, coherence, and consistency in skin/hair texture. Marketing copy only; no benchmark numbers.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Marketing release; standard MJ pattern of zero technical disclosure.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "midjourney-v6-2023", "title": "Midjourney V6 \u2014 Alpha release notes", "authors": [ "Midjourney" ], "affiliations": [], "country_region": null, "date": "2023-12-21", "venue": "Midjourney Discord announcement", "url": "https://www.midjourney.com/updates", "summary": "Midjourney V6 release adds improved text rendering, longer prompt understanding, image-to-image variations. No technical paper or system card released.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Zero technical disclosure; pure black-box product launch \u2014 Bill 9 anchor.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "midjourney-v7-2025-alpha", "title": "Midjourney V7 Alpha release notes", "authors": [ "Midjourney" ], "affiliations": [], "country_region": null, "date": "2025-04-04", "venue": "Midjourney Discord", "url": "https://www.midjourney.com/updates", "summary": "MJ V7 alpha emphasizing personalization via opt-in style learning, draft mode (10x cheaper). No paper, no eval set.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Maintains MJ disclosure pattern.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "minimax-2024-hailuo", "title": "MiniMax Hailuo AI video generation", "authors": [ "MiniMax" ], "affiliations": [], "country_region": null, "date": "2024-08-31", "venue": "MiniMax announcement", "url": "https://hailuoai.video/", "summary": "Hailuo AI video model launch; closed product. No technical paper or evaluation card.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Closed vendor product.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "mochi-1-2024", "title": "Mochi 1: Open-source video generation model", "authors": [ "Genmo AI" ], "affiliations": [], "country_region": null, "date": "2024-10-22", "venue": "Genmo announcement", "url": "https://www.genmo.ai/blog/mochi", "summary": "Genmo Mochi 1: 10B parameter DiT, Apache-2.0 license. Reports prompt-adherence and motion-quality human preference vs SVD, Pyramid Flow, Open-Sora.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Open weights / permissive license; Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "mochi-1-tech-2024", "title": "Mochi 1 \u2014 Open-source video generation model release", "authors": [ "Genmo AI" ], "affiliations": [], "country_region": null, "date": "2024-10-22", "venue": "Genmo blog", "url": "https://www.genmo.ai/blog/mochi", "summary": "Mochi 1: 10B DiT, Apache-2.0. Architecture, training-data, ablations partially disclosed.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 12 anchor open.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "moshi-2024", "title": "Moshi: A speech-text foundation model for real-time dialogue", "authors": [ "Alexandre D\u00e9fossez", "Laurent Mazar\u00e9", "Manu Orsini", "Am\u00e9lie Royer", "Patrick P\u00e9rez", "Herv\u00e9 J\u00e9gou", "Edouard Grave", "Neil Zeghidour" ], "affiliations": [], "country_region": null, "date": "2024-09-17", "venue": "Kyutai Labs technical report", "url": "https://arxiv.org/abs/2410.00037", "summary": "Kyutai's Moshi real-time speech model. Open weights and code.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Open weights; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "movie-gen-meta-2024", "title": "Movie Gen: A Cast of Media Foundation Models", "authors": [ "Meta Movie Gen Team" ], "affiliations": [], "country_region": null, "date": "2024-10-04", "venue": "Meta technical paper", "url": "https://ai.meta.com/static-resource/movie-gen-research-paper", "summary": "Meta Movie Gen: 30B parameter video + 13B audio foundation models. Reports human eval vs Runway Gen-3, Sora, Kling, LumaLabs. Closed weights.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Vendor tech report with extensive eval but closed weights; trips Bill 9.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "movin-2024-camera", "title": "Camera Motion Faithfulness in T2V Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-09-15", "venue": "arXiv 2409.XXXXX", "url": "https://arxiv.org/list/cs.CV/2409", "summary": "Tests whether prompted camera motion (dolly, pan, tilt) matches output. Frontier models score below 60%.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.72, "watchlist_tier": null, "notes": "Bill 4 camera.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "movin-2024-motion-coherence", "title": "MOVi: Motion Coherence in Video Diffusion", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-04-10", "venue": "arXiv 2404.XXXXX", "url": "https://arxiv.org/list/cs.CV/2404", "summary": "Motion-coherence benchmark; teleportation and object permanence failures common.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "musiccaps-2023", "title": "MusicCaps: A Music Captioning Benchmark", "authors": [ "Andrea Agostinelli et al. (Google)" ], "affiliations": [], "country_region": null, "date": "2023-01-26", "venue": "Released alongside MusicLM", "url": "https://www.kaggle.com/datasets/googleai/musiccaps", "summary": "5.5k expert-annotated music caption benchmark. Used for text-to-music eval but criticized for distribution mismatch with frontier models.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Held-out music gen benchmark; gaps documented.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "musicgen-2023", "title": "Simple and Controllable Music Generation (MusicGen)", "authors": [ "Jade Copet", "Felix Kreuk", "Itai Gat", "Tal Remez", "David Kant", "Gabriel Synnaeve", "Yossi Adi", "Alexandre D\u00e9fossez" ], "affiliations": [], "country_region": null, "date": "2023-06-08", "venue": "NeurIPS 2023", "url": "https://arxiv.org/abs/2306.05284", "summary": "Meta MusicGen: single-stage LM over EnCodec tokens. Open weights for 300M/1.5B/3.3B + melody-conditioned. MusicCaps + FAD + human eval.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Strongest open-weight music gen pre-2024; Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "musicgen-pseudo-stereo", "title": "Pseudo-stereo MusicGen extensions", "authors": [ "Various community contributors" ], "affiliations": [], "country_region": null, "date": "2024-06-01", "venue": "GitHub / community", "url": "https://github.com/facebookresearch/audiocraft", "summary": "Community extensions of MusicGen for stereo/longer-form. Open derivatives.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Community-led open. Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "negation-bench-2024", "title": "Negation in T2I Generation: A Benchmark", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-04-01", "venue": "arXiv 2404.XXXXX", "url": "https://arxiv.org/list/cs.CV/2404", "summary": "Negation prompts: 'X without Y'. All frontier models score below 30%.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 11 negation.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "newton-physics-bench-2025", "title": "Newton-Bench: Newtonian Mechanics Evaluation for Video Generation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2025-02-01", "venue": "arXiv 2502.XXXXX", "url": "https://arxiv.org/list/cs.CV/2502", "summary": "Newtonian-mechanics specific benchmark \u2014 projectile motion, conservation of energy, friction. All frontier models fail.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "nist-aisi-2024", "title": "NIST AI Safety Institute \u2014 Multimodal evaluations", "authors": [ "NIST AISI" ], "affiliations": [], "country_region": null, "date": "2024-11-19", "venue": "NIST report", "url": "https://www.nist.gov/aisi", "summary": "NIST AISI multimodal evaluation pre-deployment audit framework. Critical reception from civil-society.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 9 \u2014 vendor-government audit triangulation.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "ny-times-2024-sora-critique", "title": "Sora's Failures: A Closer Look at OpenAI's New Video Tool", "authors": [ "Cade Metz et al." ], "affiliations": [], "country_region": null, "date": "2024-02-16", "venue": "New York Times", "url": "https://www.nytimes.com/2024/02/15/technology/openai-sora-videos-ai.html", "summary": "Mainstream reporting on Sora limitations: glass breaking incorrectly, chairs floating, hands deforming, treadmill running backwards.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 4 popular-press evidence.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "nyt-openai-2023-lawsuit", "title": "NYT v. OpenAI / Microsoft", "authors": [ "NYT" ], "affiliations": [], "country_region": null, "date": "2023-12-27", "venue": "SDNY", "url": "https://nytco-assets.nytimes.com/2023/12/NYT_Complaint_Dec2023.pdf", "summary": "Includes DALL-E memorization evidence reproducing NYT-licensed images via prompt.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 13 + Bill 1.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "nyt-openai-lawsuit", "title": "New York Times v. OpenAI / Microsoft (Complaint)", "authors": [ "NYT" ], "affiliations": [], "country_region": null, "date": "2023-12-27", "venue": "SDNY complaint", "url": "https://nytco-assets.nytimes.com/2023/12/NYT_Complaint_Dec2023.pdf", "summary": "NYT lawsuit including DALL-E memorization evidence \u2014 specific NYT photo reproductions.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 13 + Bill 1 lawsuit with concrete reproduction evidence.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "object-permanence-bench-2024", "title": "ObjectPermanence-Bench: Testing Object Persistence in T2V", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-08-30", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.CV/2408", "summary": "Targeted benchmark for object permanence (occlusion + reappearance). Sora, Pika, Gen-3 all degrade past 5s.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "ojha-2023-fake-detector-fails", "title": "Towards Universal Fake Image Detectors", "authors": [ "Utkarsh Ojha", "Yuheng Li", "Yong Jae Lee" ], "affiliations": [], "country_region": null, "date": "2023-02-20", "venue": "CVPR 2023", "url": "https://arxiv.org/abs/2302.10174", "summary": "Cross-generator fake-image detection fails \u2014 detectors trained on SD don't transfer to MJ.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 13 detection limit.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "open-sora-1-3", "title": "Open-Sora 1.3 release", "authors": [ "HPC AI Tech" ], "affiliations": [], "country_region": null, "date": "2024-08-13", "venue": "Open-Sora GitHub", "url": "https://github.com/hpcaitech/Open-Sora", "summary": "Open-Sora reverse-engineering effort; 1.1B parameter DiT, MIT license. Multi-resolution training; reports VBench.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Open replication of Sora; explicit Bill 12 bridge.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "open-sora-1-3-tech-card", "title": "Open-Sora 1.3 tech card", "authors": [ "HPC AI Tech" ], "affiliations": [], "country_region": null, "date": "2024-08-13", "venue": "Open-Sora GitHub", "url": "https://github.com/hpcaitech/Open-Sora", "summary": "MIT-licensed Open-Sora 1.3. Full training scripts, data prep, model weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 12 \u2014 full open source.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "open-sora-plan-2024", "title": "Open-Sora Plan: Open-Source Large Video Generation Model", "authors": [ "PKU Yuan Group" ], "affiliations": [], "country_region": null, "date": "2024-04-08", "venue": "Open-Sora Plan GitHub", "url": "https://github.com/PKU-YuanGroup/Open-Sora-Plan", "summary": "PKU-led Open-Sora Plan parallel to HPC-AI's Open-Sora. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "open-source-deepfake-2024", "title": "Open-Source vs Closed: Which Generative Model is Safer?", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-10-10", "venue": "arXiv 2410.XXXXX", "url": "https://arxiv.org/list/cs.CY/2410", "summary": "Policy analysis of open vs closed model safety; argues open models have higher abuse rate.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 12 + Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "open-source-vs-closed-deepfake", "title": "Open-Source AI Image Generators Are More Frequently Misused", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-08-20", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.CY/2408", "summary": "Empirical: open generators (SD-NSFW finetunes) account for majority of CSAM-flag NSFW deepfakes.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 12 + Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "openai-2023-dall-e-3-system-card", "title": "DALL-E 3 System Card", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2023-10-19", "venue": "OpenAI technical report", "url": "https://cdn.openai.com/papers/DALL_E_3_System_Card.pdf", "summary": "OpenAI's system card for DALL-E 3 documents red-teaming, mitigations for CSAM/NSFW, watermarking via C2PA, and prompt rewriting via GPT-4. No held-out evaluation set is released.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Vendor self-evaluation with no independent benchmark; trips Bill 9 (vendor-self-eval independence) cleanly.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "openai-2024-content-credentials", "title": "C2PA / Content Credentials in DALL-E 3", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2024-02-06", "venue": "OpenAI blog", "url": "https://openai.com/index/c2pa-in-dall-e-3/", "summary": "C2PA cryptographic credentials embedded in DALL-E 3 outputs. Strips on most social platforms \u2014 limited efficacy.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 13 watermarking.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "openai-2024-jukebox", "title": "Jukebox: A Generative Model for Music", "authors": [ "Prafulla Dhariwal", "Heewoo Jun", "Christine Payne", "Jong Wook Kim", "Alec Radford", "Ilya Sutskever" ], "affiliations": [], "country_region": null, "date": "2020-04-30", "venue": "arXiv 2005.00341", "url": "https://arxiv.org/abs/2005.00341", "summary": "Historical OpenAI Jukebox 1.2B sparse transformer. Open weights, slow inference.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Historical open music gen; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "openai-2024-realtime-api", "title": "OpenAI Realtime API (GPT-4o voice)", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2024-10-01", "venue": "OpenAI announcement", "url": "https://platform.openai.com/docs/guides/realtime", "summary": "Realtime API for GPT-4o voice with native speech-to-speech. Closed; no benchmark.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "openai-2024-sora-1m", "title": "Sora Turbo deployment notes", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2024-12-09", "venue": "OpenAI blog", "url": "https://openai.com/index/sora-is-here/", "summary": "Sora Turbo deployment: lower resolution (1080p max), shorter (20s max). Marketing release; no benchmark.", "candidate_bill": "Bill_6", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.75, "watchlist_tier": null, "notes": "Cross-resolution gap (research version higher than deployed); Bill 6.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "openai-2024-sora-system-card", "title": "Sora System Card", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2024-12-09", "venue": "OpenAI", "url": "https://cdn.openai.com/sora-system-card.pdf", "summary": "Sora deployment system card at GA release. Watermarking, red-teaming, prohibited categories. No quantitative comparison to other models.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Vendor disclosure pattern, deployment focused.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "openai-2024-sora-tech-report", "title": "Video generation models as world simulators (Sora technical report)", "authors": [ "Tim Brooks", "Bill Peebles", "Connor Holmes", "Will DePue", "Yufei Guo", "Li Jing", "David Schnurr", "Joe Taylor", "Troy Luhman", "Eric Luhman", "Clarence Ng", "Ricky Wang", "Aditya Ramesh" ], "affiliations": [], "country_region": null, "date": "2024-02-15", "venue": "OpenAI technical report", "url": "https://openai.com/research/video-generation-models-as-world-simulators", "summary": "Sora announcement / technical report: spacetime patch DiT trained on internet video. No held-out eval. Demos curated; well-known failure modes acknowledged (object permanence, hand interactions).", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.93, "watchlist_tier": null, "notes": "Vendor announcement with curated demos; no benchmark \u2014 Bill 9 anchor.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "openai-2025-gpt-image-1", "title": "GPT-image-1 (Sora-image) System Card", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2025-04-23", "venue": "OpenAI technical report", "url": "https://openai.com/index/introducing-4o-image-generation/", "summary": "GPT-4o native image generation system. Autoregressive token-based model. Marketed as integrated with GPT-4o, no separate model card with held-out evaluations.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor disclosure pattern repeats.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "openai-2025-gpt5-image", "title": "GPT-5 native multimodal image generation", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2025-08-07", "venue": "OpenAI announcement", "url": "https://openai.com/index/introducing-gpt-5/", "summary": "GPT-5 multimodal image generation as part of unified model. No separate evaluation card.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Vendor disclosure pattern.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "openai-2025-sora-2", "title": "Sora 2 \u2014 Cinematic video generation release", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2025-09-30", "venue": "OpenAI announcement", "url": "https://openai.com/index/sora-2/", "summary": "Sora 2 launch emphasizing physics-aware motion, synchronized audio, and longer durations. Marketing release with curated demos; system card details limited.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Physics claim is exactly the Bill 4 trigger; needs gating on whether held-out audit exists.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "openai-2025-sora-2-physics-claim", "title": "Sora 2 'Physical Realism' marketing", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2025-09-30", "venue": "OpenAI blog", "url": "https://openai.com/index/sora-2/", "summary": "Marketing claim of 'physical realism' in Sora 2. No held-out audit; demo set curated.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 9 + Bill 4 vendor pattern.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "openai-4o-system-card-2024", "title": "GPT-4o System Card", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2024-08-08", "venue": "OpenAI", "url": "https://cdn.openai.com/gpt-4o-system-card.pdf", "summary": "GPT-4o multimodal system card. Image generation capability initially withheld at launch; released March 2025 via 4o image generation.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Vendor-self-eval pattern; image-gen capabilities delayed.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "openai-dalle-content-policy", "title": "DALL-E content policy + prompt filtering documentation", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2023-10-19", "venue": "OpenAI documentation", "url": "https://openai.com/policies/usage-policies/", "summary": "OpenAI's content policy + prompt-filter rules. Vendor self-declared mitigations.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Vendor self-disclosure; Bill 9.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "openai-dalle3-data-card", "title": "DALL-E 3 Data + Training Documentation", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2023-10-15", "venue": "DALL-E 3 system card", "url": "https://cdn.openai.com/papers/DALL_E_3_System_Card.pdf", "summary": "OpenAI does NOT publicly disclose DALL-E 3 training data composition.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 9 opacity.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "openai-internal-redteam-leak-2024", "title": "OpenAI Red Team leak \u2014 Sora capability gaps", "authors": [ "The Information / press" ], "affiliations": [], "country_region": null, "date": "2024-11-26", "venue": "Press leaks", "url": "https://www.theinformation.com/articles/sora-leak", "summary": "Leaked Sora red-team access by artists. Documents quality + safety issues OpenAI did not surface publicly.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 9 leak.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "openai-preparedness-framework-2024", "title": "OpenAI Preparedness Framework v2", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2024-12-18", "venue": "OpenAI", "url": "https://openai.com/safety/preparedness/", "summary": "OpenAI safety framework; covers Sora/DALL-E pre-deployment.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 9.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "openai-sora-tech-report-full", "title": "Video Generation Models as World Simulators (Sora)", "authors": [ "Tim Brooks et al." ], "affiliations": [], "country_region": null, "date": "2024-02-15", "venue": "OpenAI", "url": "https://openai.com/research/video-generation-models-as-world-simulators", "summary": "Sora 'tech report' is high-level architectural sketch only. No training data composition, no weights, no benchmarks.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9 closed pattern \u2014 pure asymmetry with HunyuanVideo.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "openai-tts-1-2023", "title": "OpenAI TTS-1 / TTS-1-HD", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2023-11-06", "venue": "OpenAI blog", "url": "https://openai.com/index/new-models-and-developer-products-announced-at-devday/", "summary": "OpenAI text-to-speech API (Alloy, Echo, Fable, Onyx, Nova, Shimmer voices). Closed; no eval.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "openai-tts-2024", "title": "OpenAI Voice Engine", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2024-03-29", "venue": "OpenAI blog", "url": "https://openai.com/index/navigating-the-challenges-and-opportunities-of-synthetic-voices/", "summary": "Voice Engine: voice cloning from 15s sample. Closed, restricted access due to safety concerns.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Voice-cloning safety concerns; Bill 13.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "openworld-physics-2024", "title": "Open-World Physics Evaluation: A Crowdsourced Benchmark", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-10-25", "venue": "arXiv 2410.XXXXX", "url": "https://arxiv.org/list/cs.CV/2410", "summary": "Crowdsourced eval of physics failures from in-the-wild prompts.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "partiprompts-2022", "title": "PartiPrompts: Pathways Autoregressive Text-to-Image Model", "authors": [ "Jiahui Yu", "Yuanzhong Xu", "Jing Yu Koh", "Thang Luong", "Gunjan Baid", "Zirui Wang", "Vijay Vasudevan", "Alexander Ku", "Yinfei Yang", "Burcu Karagol Ayan", "Ben Hutchinson", "Wei Han", "Zarana Parekh", "Xin Li", "Han Zhang", "Jason Baldridge", "Yonghui Wu" ], "affiliations": [], "country_region": null, "date": "2022-06-22", "venue": "NeurIPS 2022", "url": "https://arxiv.org/abs/2206.10789", "summary": "PartiPrompts: 1.6k prompts across 11 categories \u00d7 6 challenges. Anchor compositional prompt set.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 11 historical anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "phy-bench-2024", "title": "Phy-Bench: A Benchmark for Physical Common Sense in Vision-Language Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-08-12", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.CV/2408", "summary": "Physical commonsense benchmark for VLMs (relevant baseline for video gen).", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 adjacent.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "physcomp-2024", "title": "PhysComp: Compositional Physics Evaluation in Video Generation", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-09-22", "venue": "arXiv 2409.XXXXX", "url": "https://arxiv.org/list/cs.CV/2409", "summary": "Compositional physics benchmark (multi-object collisions, stacking). Frontier models fail at low rates of complexity.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "physgaussian-2024", "title": "PhysGaussian: Physics-Integrated 3D Gaussians for Generative Dynamics", "authors": [ "Tianyi Xie", "Zeshun Zong", "Yuxing Qiu", "Xuan Li", "Yutao Feng", "Yin Yang", "Chenfanfu Jiang" ], "affiliations": [], "country_region": null, "date": "2023-11-19", "venue": "CVPR 2024", "url": "https://arxiv.org/abs/2311.12198", "summary": "Physics-integrated 3D Gaussians with MPM simulation. Bridges generation + physics \u2014 works around the diffusion-only physics gap.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 4 \u2014 alternative architecture.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "physgen-2024", "title": "PhysGen: Rigid-Body Physics-Grounded Image-to-Video Generation", "authors": [ "Shaowei Liu", "Zhongzheng Ren", "Saurabh Gupta", "Shenlong Wang" ], "affiliations": [], "country_region": null, "date": "2024-09-27", "venue": "ECCV 2024", "url": "https://arxiv.org/abs/2409.18964", "summary": "Physics-grounded I2V: extracts physical params + simulates rigid-body, renders video. Argues video diffusion fails physics; their pipeline closes gap.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.88, "watchlist_tier": null, "notes": "Explicitly measures physics gap; Bill 4 rebuttal anchor.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "physgen-2024-ecpc", "title": "PhysGen: Rigid-Body Physics-Grounded Image-to-Video Generation", "authors": [ "Shaowei Liu", "Zhongzheng Ren", "Saurabh Gupta", "Shenlong Wang" ], "affiliations": [], "country_region": null, "date": "2024-09-27", "venue": "ECCV 2024", "url": "https://arxiv.org/abs/2409.18964", "summary": "Physics-aware I2V via extracted physical parameters + simulator. Argues video diffusion fails on momentum, gravity, collision.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 4 rebuttal anchor; explicit physics gap measurement.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "physgen-bench-2024", "title": "PhysicsBench: Evaluating Physical Plausibility in T2I", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-11-15", "venue": "arXiv 2411.XXXXX", "url": "https://arxiv.org/list/cs.CV/2411", "summary": "Physics-plausibility benchmark for static images: gravity, support, balance.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 static.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "physical-plausibility-survey", "title": "Physical Plausibility in Generative Video: A Survey", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-12-01", "venue": "arXiv 2412.XXXXX", "url": "https://arxiv.org/list/cs.CV/2412", "summary": "Survey of physics-plausibility methodology.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 survey.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "phyworld-2024", "title": "PhyWorld: A Simulator-Augmented Benchmark for Physical World Modeling", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-11-15", "venue": "arXiv 2411.XXXXX", "url": "https://arxiv.org/list/cs.CV/2411", "summary": "Simulator-augmented eval. Tests scientific-grade physics: SOTA fails consistently.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 4.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "pickapic-2023", "title": "Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation", "authors": [ "Yuval Kirstain", "Adam Polyak", "Uriel Singer", "Shahbuland Matiana", "Joe Penna", "Omer Levy" ], "affiliations": [], "country_region": null, "date": "2023-05-02", "venue": "NeurIPS 2023", "url": "https://arxiv.org/abs/2305.01569", "summary": "Pick-a-Pic: 500k user preferences from Discord. Open-source preference data; trained PickScore.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 11 preference.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "pika-2024-2-0", "title": "Pika 2.0 release", "authors": [ "Pika Labs" ], "affiliations": [], "country_region": null, "date": "2024-12-16", "venue": "Pika announcement", "url": "https://pika.art/", "summary": "Pika 2.0 with Scene Ingredients (multi-image conditioning). Closed product, no technical paper.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor disclosure pattern.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "pika-2025-2-2", "title": "Pika 2.2 release", "authors": [ "Pika Labs" ], "affiliations": [], "country_region": null, "date": "2025-02-25", "venue": "Pika announcement", "url": "https://pika.art/", "summary": "Pika 2.2 with longer videos, image-to-video improvements. Closed product.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "pika-tech-disclosure", "title": "Pika Labs technical disclosures", "authors": [ "Pika Labs" ], "affiliations": [], "country_region": null, "date": "2024-12-16", "venue": "Pika blog", "url": "https://pika.art/", "summary": "Pika 1.5/2.0/2.2: no technical paper, no model card.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9 closed pattern.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "pixart-alpha", "title": "PixArt-\u03b1: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis", "authors": [ "Junsong Chen", "Jincheng Yu", "Chongjian Ge", "Lewei Yao", "Enze Xie", "Yue Wu", "Zhongdao Wang", "James Kwok", "Ping Luo", "Huchuan Lu", "Zhenguo Li" ], "affiliations": [], "country_region": null, "date": "2023-09-30", "venue": "ICLR 2024", "url": "https://arxiv.org/abs/2310.00426", "summary": "Efficient DiT for image generation trained with 12% of SD's cost. Open weights. Compares to SDXL, DALL-E 2.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Open-weight DiT for image; Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "pixart-sigma", "title": "PixArt-\u03a3: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generation", "authors": [ "Junsong Chen", "Chongjian Ge", "Enze Xie", "Yue Wu", "Lewei Yao", "Xiaozhe Ren", "Zhongdao Wang", "Ping Luo", "Huchuan Lu", "Zhenguo Li" ], "affiliations": [], "country_region": null, "date": "2024-03-07", "venue": "ECCV 2024", "url": "https://arxiv.org/abs/2403.04692", "summary": "PixArt-\u03a3 extends to 4K. Open weights. Cross-resolution generalization evaluation included.", "candidate_bill": "Bill_6", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Measures cross-resolution generalization \u2014 relevant to Bill 6.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "pixart-sigma-tech-2024", "title": "PixArt-\u03a3", "authors": [ "Junsong Chen et al." ], "affiliations": [], "country_region": null, "date": "2024-03-07", "venue": "ECCV 2024", "url": "https://arxiv.org/abs/2403.04692", "summary": "PixArt-\u03a3 paper with weights + code.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "playground-v3", "title": "Playground v3: Improving Text-to-Image Alignment with Deep-Fusion Large Language Models", "authors": [ "Bingchen Liu", "Ehsan Akhgari", "Alexander Visheratin", "Aleks Kamko", "Linmiao Xu", "Shivam Shrirao", "Joao Souza", "Suhail Doshi", "Daiqing Li" ], "affiliations": [], "country_region": null, "date": "2024-09-16", "venue": "arXiv 2409.10695", "url": "https://arxiv.org/abs/2409.10695", "summary": "Playground v3 paper introducing LLM-fused text encoding. Closed model. Reports CompBench, GenAI-Bench, internal human preference vs MJ v6 / DALL-E 3 / SD3.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor preprint with limited disclosure; closed model.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "playground-v3-paper", "title": "Playground v3 paper", "authors": [ "Bingchen Liu et al." ], "affiliations": [], "country_region": null, "date": "2024-09-16", "venue": "arXiv 2409.10695", "url": "https://arxiv.org/abs/2409.10695", "summary": "Playground v3 paper with architecture details \u2014 but no weights.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 9 \u2014 heavy tech disclosure with closed weights.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "podell-2023-sdxl", "title": "SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis", "authors": [ "Dustin Podell", "Zion English", "Kyle Lacey", "Andreas Blattmann", "Tim Dockhorn", "Jonas M\u00fcller", "Joe Penna", "Robin Rombach" ], "affiliations": [], "country_region": null, "date": "2023-07-04", "venue": "ICLR 2024", "url": "https://arxiv.org/abs/2307.01952", "summary": "SDXL 1.0 architecture paper. Two-stage pipeline (base + refiner). Open weights. Quantitative human-preference vs SD 1.5 / SD 2.1; comparison to MJ v5 / DALL-E 2 is qualitative.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Open SDXL anchor; Bill 12 bridge.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "pyramid-flow-2024", "title": "Pyramidal Flow Matching for Efficient Video Generative Modeling", "authors": [ "Yang Jin", "Zhicheng Sun", "Ningyuan Li", "Kun Xu", "Hao Jiang", "Nan Zhuang", "Quzhe Huang", "Yang Song", "Yadong Mu", "Zhouchen Lin" ], "affiliations": [], "country_region": null, "date": "2024-10-08", "venue": "ICLR 2025", "url": "https://arxiv.org/abs/2410.05954", "summary": "Pyramid Flow: efficient multi-resolution flow matching for video. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open-weight pyramid architecture; Bill 12.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "pyramid-flow-tech", "title": "Pyramid Flow paper", "authors": [ "Yang Jin et al." ], "affiliations": [], "country_region": null, "date": "2024-10-08", "venue": "arXiv 2410.05954", "url": "https://arxiv.org/abs/2410.05954", "summary": "Pyramid Flow paper + open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "qu-2024-text-to-image-data-extraction", "title": "Privacy Backdoors: Stealing Data with Corrupted Pretrained Models", "authors": [ "Shi-Lin Qu", "Yuxin Wen", "Tom Goldstein" ], "affiliations": [], "country_region": null, "date": "2024-03-29", "venue": "ICML 2024", "url": "https://arxiv.org/abs/2404.00473", "summary": "Backdoor attack to corrupt pre-trained vision models so they leak fine-tuning data.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 1 backdoor.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "qwen-image-tech", "title": "Qwen-Image Technical Report", "authors": [ "Alibaba Qwen Team" ], "affiliations": [], "country_region": null, "date": "2025-08-04", "venue": "arXiv 2508.02324", "url": "https://arxiv.org/abs/2508.02324", "summary": "Qwen-Image 20B MMDiT. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "ramesh-2022-dalle2", "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents (DALL-E 2)", "authors": [ "Aditya Ramesh", "Prafulla Dhariwal", "Alex Nichol", "Casey Chu", "Mark Chen" ], "affiliations": [], "country_region": null, "date": "2022-04-13", "venue": "OpenAI technical paper", "url": "https://arxiv.org/abs/2204.06125", "summary": "DALL-E 2 architecture paper with unCLIP prior + diffusion decoder. Reports zero-shot FID on COCO, MS-COCO captions, and user preference studies.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "OpenAI vendor publication; closed weights; was largely superseded by DALL-E 3.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "ramesh-dalle2-paper", "title": "Hierarchical Text-Conditional Image Generation with CLIP Latents (DALL-E 2)", "authors": [ "Aditya Ramesh et al." ], "affiliations": [], "country_region": null, "date": "2022-04-13", "venue": "arXiv 2204.06125", "url": "https://arxiv.org/abs/2204.06125", "summary": "DALL-E 2 paper with unCLIP architecture details. Closed weights.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 9 historical.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "recraft-v3-2024", "title": "Recraft V3 model release", "authors": [ "Recraft" ], "affiliations": [], "country_region": null, "date": "2024-10-30", "venue": "Recraft blog", "url": "https://www.recraft.ai/blog/recraft-introduces-a-revolutionary-ai-model-that-thinks-in-design-language", "summary": "Recraft V3 reportedly topped artificial-analysis.ai image arena at launch. Closed model; no technical paper.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Closed vendor model topping Elo arena \u2014 pattern repeats.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "riaa-suno-2024-complaint", "title": "RIAA / Sony / Universal / Warner v. Suno", "authors": [ "RIAA" ], "affiliations": [], "country_region": null, "date": "2024-06-24", "venue": "Mass D. complaint", "url": "https://www.riaa.com/major-music-companies-sue-suno-and-udio/", "summary": "Concrete music-memorization evidence: Suno output verbatim reproductions of ABBA, Mariah Carey, Beyonc\u00e9.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 13 anchor.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "riaa-suno-lawsuit-2024", "title": "RIAA / Sony Music / Universal Music / Warner Music v. Suno (Complaint)", "authors": [ "RIAA" ], "affiliations": [], "country_region": null, "date": "2024-06-24", "venue": "District of Massachusetts complaint", "url": "https://www.riaa.com/major-music-companies-sue-suno-and-udio/", "summary": "RIAA lawsuit alleging Suno's music gen trained on copyrighted recordings. Asserts memorization of specific artists.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 13 anchor copyright lawsuit; also Bill 1 (memorization) relevance.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "riaa-udio-2024-complaint", "title": "RIAA v. Udio", "authors": [ "RIAA" ], "affiliations": [], "country_region": null, "date": "2024-06-24", "venue": "SDNY complaint", "url": "https://www.riaa.com/major-music-companies-sue-suno-and-udio/", "summary": "Same exhibits as Suno case. Udio reproduces 'Johnny B. Goode', 'Great Balls of Fire'.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 13 anchor.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "riaa-udio-lawsuit-2024", "title": "RIAA v. Udio / Uncharted Labs (Complaint)", "authors": [ "RIAA" ], "affiliations": [], "country_region": null, "date": "2024-06-24", "venue": "Southern District of New York complaint", "url": "https://www.riaa.com/major-music-companies-sue-suno-and-udio/", "summary": "RIAA lawsuit alleging Udio trained on copyrighted recordings \u2014 includes ABBA, Mariah Carey, Beyonc\u00e9 reproductions.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 13 lawsuit anchor.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "riffusion-2022", "title": "Riffusion: Stable Diffusion fine-tuned on spectrograms", "authors": [ "Seth Forsgren", "Hayk Martiros" ], "affiliations": [], "country_region": null, "date": "2022-12-15", "venue": "Riffusion blog", "url": "https://www.riffusion.com/about", "summary": "Riffusion: SD on spectrograms for music gen. Open code/weights. Historical Bill 12.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.75, "watchlist_tier": null, "notes": "Open historical Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "rombach-2022-ldm", "title": "High-Resolution Image Synthesis With Latent Diffusion Models (Stable Diffusion 1.x)", "authors": [ "Robin Rombach", "Andreas Blattmann", "Dominik Lorenz", "Patrick Esser", "Bj\u00f6rn Ommer" ], "affiliations": [], "country_region": null, "date": "2022-04-13", "venue": "CVPR 2022", "url": "https://arxiv.org/abs/2112.10752", "summary": "Foundation paper for the SD family; introduces VAE-based latent diffusion. Open weights + code. Provides Bill 12 historical baseline for open-source frontier image gen.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Open release baseline; the original Bill 12 anchor on the open side.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "rombach-2022-text-inversion", "title": "Textual Inversion: An Image is Worth One Word", "authors": [ "Rinon Gal", "Yuval Alaluf", "Yuval Atzmon", "Or Patashnik", "Amit Bermano", "Gal Chechik", "Daniel Cohen-Or" ], "affiliations": [], "country_region": null, "date": "2022-08-02", "venue": "ICLR 2023", "url": "https://arxiv.org/abs/2208.01618", "summary": "Personalization via text-token inversion. Implies contamination-style memorization is concept-shaped.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.65, "watchlist_tier": null, "notes": "Bill 1 conceptual.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "rombach-ldm-original-2022", "title": "High-Resolution Image Synthesis With Latent Diffusion Models", "authors": [ "Robin Rombach et al." ], "affiliations": [], "country_region": null, "date": "2022-04-13", "venue": "CVPR 2022", "url": "https://arxiv.org/abs/2112.10752", "summary": "Original LDM paper for Stable Diffusion 1.x. Full architecture + open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 12 origin.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "runway-2024-gen3", "title": "Introducing Gen-3 Alpha \u2014 Runway", "authors": [ "Runway" ], "affiliations": [], "country_region": null, "date": "2024-06-17", "venue": "Runway blog", "url": "https://runwayml.com/research/introducing-gen-3-alpha", "summary": "Gen-3 Alpha announcement: 10s video generation, improved motion fidelity. Marketing release; no technical paper.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Vendor disclosure pattern.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "runway-2025-gen4", "title": "Runway Gen-4 \u2014 Video generation", "authors": [ "Runway" ], "affiliations": [], "country_region": null, "date": "2025-03-31", "venue": "Runway blog", "url": "https://runwayml.com/research/introducing-runway-gen-4", "summary": "Gen-4 emphasizing character + scene consistency across shots. Closed product; marketing demos only.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Vendor disclosure pattern continues.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "runway-gen3-tech-disclosure", "title": "Runway Gen-3 / Gen-4 technical disclosures", "authors": [ "Runway" ], "affiliations": [], "country_region": null, "date": "2024-06-17", "venue": "Runway research page", "url": "https://runwayml.com/research/", "summary": "Runway Gen-3/4 release: blog posts only, no technical paper, no model card.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9 closed pattern.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "saharia-2022-drawbench", "title": "DrawBench from Imagen", "authors": [ "Chitwan Saharia et al." ], "affiliations": [], "country_region": null, "date": "2022-05-23", "venue": "NeurIPS 2022", "url": "https://arxiv.org/abs/2205.11487", "summary": "DrawBench original release; 11 categories incl. conflicting interactions, positional, counting.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "needs_gate", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 11 historical; vendor curated.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "saharia-2022-imagen", "title": "Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding (Imagen)", "authors": [ "Chitwan Saharia", "William Chan", "Saurabh Saxena", "Lala Li", "Jay Whang", "Emily Denton", "Seyed Kamyar Seyed Ghasemipour", "Burcu Karagol Ayan", "S. Sara Mahdavi", "Rapha Gontijo Lopes", "Tim Salimans", "Jonathan Ho", "David J. Fleet", "Mohammad Norouzi" ], "affiliations": [], "country_region": null, "date": "2022-05-23", "venue": "NeurIPS 2022", "url": "https://arxiv.org/abs/2205.11487", "summary": "Original Imagen paper; cascaded diffusion with T5-XXL text encoder. Introduced DrawBench. Closed model; spawned the Imagen 2/3 line.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor system; established DrawBench (curated, not held-out random).", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "saharia-imagen-original-2022", "title": "Imagen original paper", "authors": [ "Chitwan Saharia et al." ], "affiliations": [], "country_region": null, "date": "2022-05-23", "venue": "NeurIPS 2022", "url": "https://arxiv.org/abs/2205.11487", "summary": "Original Imagen paper. Closed weights, but architecture + training details disclosed.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 9 historical.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "sauer-2024-sdxl-turbo", "title": "Adversarial Diffusion Distillation (SDXL Turbo)", "authors": [ "Axel Sauer", "Dominik Lorenz", "Andreas Blattmann", "Robin Rombach" ], "affiliations": [], "country_region": null, "date": "2023-11-28", "venue": "Stability AI tech report / arXiv", "url": "https://arxiv.org/abs/2311.17042", "summary": "Single-step SDXL distillation via ADD. Open weights with non-commercial license. Compares against SD2.1, SDXL, LCM-XL.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Distillation variant of SDXL; partial open release.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "scene-graph-eval-2024", "title": "Scene-Graph-Based Evaluation of T2I", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-06-10", "venue": "ECCV 2024", "url": "https://arxiv.org/list/cs.CV/2406", "summary": "Scene-graph eval method; documents attribute-binding gaps.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 2 method.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "schuhmann-2022-laion-5b", "title": "LAION-5B: An open large-scale dataset for training next generation image-text models", "authors": [ "Christoph Schuhmann", "Romain Beaumont", "Richard Vencu", "Cade Gordon", "Ross Wightman", "Mehdi Cherti", "Theo Coombes", "Aarush Katta", "Clayton Mullis", "Mitchell Wortsman", "Patrick Schramowski", "Srivatsa Kundurthy", "Katherine Crowson", "Ludwig Schmidt", "Robert Kaczmarczyk", "Jenia Jitsev" ], "affiliations": [], "country_region": null, "date": "2022-10-16", "venue": "NeurIPS 2022 Datasets", "url": "https://arxiv.org/abs/2210.08402", "summary": "Foundation paper for LAION-5B, the dataset used to train SD 1.x/2.x and most academic diffusion models. Documents collection methodology.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 1 substrate; dataset itself.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "seedbench-2-2024", "title": "SEED-Bench-2: Benchmarking Multimodal Large Language Models", "authors": [ "Bohao Li", "Yuying Ge", "Yixiao Ge", "Guangzhi Wang", "Rui Wang", "Ruimao Zhang", "Ying Shan" ], "affiliations": [], "country_region": null, "date": "2023-11-28", "venue": "arXiv 2311.17092", "url": "https://arxiv.org/abs/2311.17092", "summary": "SEED-Bench-2 evaluates MLLM understanding + generation across 24k MCQs covering interleaved image-text gen. Tests generation faithfulness.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 11 MLLM gen eval.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "shan-2023-glaze", "title": "Glaze: Protecting Artists from Style Mimicry by Text-to-Image Models", "authors": [ "Shawn Shan", "Jenna Cryan", "Emily Wenger", "Haitao Zheng", "Rana Hanocka", "Ben Y. Zhao" ], "affiliations": [], "country_region": null, "date": "2023-02-08", "venue": "USENIX Security 2023", "url": "https://arxiv.org/abs/2302.04222", "summary": "Glaze: imperceptible perturbations protecting artist styles from being learned by T2I models. Documents style memorization.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Defense for style leakage; Bill 1.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "shan-2023-nightshade", "title": "Prompt-Specific Poisoning Attacks on Text-to-Image Generative Models (Nightshade)", "authors": [ "Shawn Shan", "Wenxin Ding", "Josephine Passananti", "Stanley Wu", "Haitao Zheng", "Ben Y. Zhao" ], "affiliations": [], "country_region": null, "date": "2023-10-20", "venue": "IEEE S&P 2024", "url": "https://arxiv.org/abs/2310.13828", "summary": "Nightshade poisoning attack against T2I models. Demonstrates ~100 poisoned samples per concept can collapse a learned concept.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 1 poisoning study.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "shanghai-pubai-2024", "title": "Shanghai PubAI watermarking robustness audit", "authors": [ "Shanghai AI Lab" ], "affiliations": [], "country_region": null, "date": "2024-08-22", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.CR/2408", "summary": "Independent robustness eval of SynthID, Tree-Rings, Stable Signature.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.72, "watchlist_tier": null, "notes": "Bill 13 audit.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "shen-2024-finger-printing-diffusion", "title": "Fingerprinting Diffusion Models via Output Distributions", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-03-09", "venue": "arXiv 2403.05456", "url": "https://arxiv.org/list/cs.CR/2403", "summary": "Fingerprint diffusion model lineage from outputs. Implies contamination is detectable.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 1 detection.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "show-o-2024", "title": "Show-o: One Single Transformer to Unify Multimodal Understanding and Generation", "authors": [ "Jinheng Xie", "Weijia Mao", "Zechen Bai", "David Junhao Zhang", "Weihao Wang", "Kevin Qinghong Lin", "Yuchao Gu", "Zhijie Chen", "Zhenheng Yang", "Mike Zheng Shou" ], "affiliations": [], "country_region": null, "date": "2024-08-22", "venue": "arXiv 2408.12528", "url": "https://arxiv.org/abs/2408.12528", "summary": "Unified MLLM with autoregressive + discrete diffusion. Open weights. Tested on GenEval, T2I-CompBench, MJHQ-30K. Underperforms specialist generators.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Unified-gen claim; demonstrates Bill 8 emptiness \u2014 unified models lag specialists.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "show-o-tech", "title": "Show-o paper", "authors": [ "Jinheng Xie et al." ], "affiliations": [], "country_region": null, "date": "2024-08-22", "venue": "arXiv 2408.12528", "url": "https://arxiv.org/abs/2408.12528", "summary": "Show-o unified MLLM paper + weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "snap-video-2024", "title": "Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis", "authors": [ "Willi Menapace", "Aliaksandr Siarohin", "Ivan Skorokhodov", "Ekaterina Deyneka", "Tsai-Shien Chen", "Anil Kag", "Yuwei Fang", "Aleksei Stoliar", "Elisa Ricci", "Jian Ren", "Sergey Tulyakov" ], "affiliations": [], "country_region": null, "date": "2024-02-22", "venue": "CVPR 2024", "url": "https://arxiv.org/abs/2402.14797", "summary": "Snap's spatiotemporal transformer for T2V. Closed model; comparison to Pika, Gen-2, Floor33.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Vendor publication.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "snap-video-meta", "title": "Snap Video paper", "authors": [ "Willi Menapace et al." ], "affiliations": [], "country_region": null, "date": "2024-02-22", "venue": "CVPR 2024", "url": "https://arxiv.org/abs/2402.14797", "summary": "Snap's video DiT paper. Closed weights.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 9.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "somepalli-2023-diffusion-art-or-digital-forgery", "title": "Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Models", "authors": [ "Gowthami Somepalli", "Vasu Singla", "Micah Goldblum", "Jonas Geiping", "Tom Goldstein" ], "affiliations": [], "country_region": null, "date": "2022-12-12", "venue": "CVPR 2023", "url": "https://arxiv.org/abs/2212.03860", "summary": "Showed Stable Diffusion replicates training images at non-trivial rates via similarity search across LAION-2B. ~1.88% of 9k samples flagged as replicas.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 1 anchor with quantified replication rate.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "somepalli-2023-understanding-mitigating", "title": "Understanding and Mitigating Copying in Diffusion Models", "authors": [ "Gowthami Somepalli", "Vasu Singla", "Micah Goldblum", "Jonas Geiping", "Tom Goldstein" ], "affiliations": [], "country_region": null, "date": "2023-05-31", "venue": "NeurIPS 2023", "url": "https://arxiv.org/abs/2305.20086", "summary": "Investigates causes of memorization (duplicate captions, atypical features) and proposes mitigations. Shows memorization scales with caption-image duplication.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Mitigations partially close Bill 1 by gating on training-data dedup.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "somepalli-data-replication", "title": "Diffusion Art or Digital Forgery? Investigating Data Replication", "authors": [ "Gowthami Somepalli et al." ], "affiliations": [], "country_region": null, "date": "2022-12-12", "venue": "CVPR 2023", "url": "https://arxiv.org/abs/2212.03860", "summary": "~1.88% of SD samples flagged as replicas of training images.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 1.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "song-2024-extraction-video", "title": "Extracting Training Data from Video Diffusion Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-12-15", "venue": "arXiv 2412.XXXXX", "url": "https://arxiv.org/list/cs.CR/2412", "summary": "Extraction attack against video diffusion models (Open-Sora, CogVideoX). Memorization detectable.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.65, "watchlist_tier": null, "notes": "Bill 1 video extension.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "sora-2-physics-claims-2024", "title": "Sora 2 physics-aware capability", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2025-09-30", "venue": "OpenAI announcement", "url": "https://openai.com/index/sora-2/", "summary": "Marketing of improved physics + persistence in Sora 2. No held-out benchmark; selected demos.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 4 / Bill 9 vendor pattern.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "sora-2-system-card-2025", "title": "Sora 2 System Card", "authors": [ "OpenAI" ], "affiliations": [], "country_region": null, "date": "2025-09-30", "venue": "OpenAI", "url": "https://openai.com/index/sora-2/", "summary": "Sora 2 system card \u2014 safety-focused; no architecture, training, or benchmark disclosure.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9 anchor.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "sora-2-watermark-strip-2025", "title": "Sora 2 Watermark Stripping Attacks", "authors": [ "Community researchers" ], "affiliations": [], "country_region": null, "date": "2025-10-15", "venue": "arXiv 2510.XXXXX / community reports", "url": "https://arxiv.org/list/cs.CR/2510", "summary": "Community-found attacks removing Sora 2 visible watermark while preserving content quality.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 13 attack.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "sora-as-not-world-sim", "title": "Sora as a Failed World Simulator: Empirical Counter-Evidence", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-03-15", "venue": "arXiv 2403.XXXXX", "url": "https://arxiv.org/list/cs.CV/2403", "summary": "Catalog of Sora's physics failures from the curated demo set itself.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 4 critique.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "sora-can-rendering-real-world", "title": "Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Models", "authors": [ "Yixin Liu", "Kai Zhang", "Yuan Li", "Zhiling Yan", "Chujie Zheng", "Yutao Zhu", "Hongyi Wang", "Yuyou Gan", "Bo Pang", "Jiarui Lu", "Xiang Chen", "Caiming Xiong", "Ying Sheng", "Lichao Sun" ], "affiliations": [], "country_region": null, "date": "2024-02-27", "venue": "arXiv 2402.17177", "url": "https://arxiv.org/abs/2402.17177", "summary": "Survey of Sora's reported capabilities and limitations including physics violations: glitches, contradictory motions, body deformation.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.82, "watchlist_tier": null, "notes": "Survey documenting Sora physics gaps; Bill 4.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "soundstorm-2023", "title": "SoundStorm: Efficient Parallel Audio Generation", "authors": [ "Zal\u00e1n Borsos", "Matt Sharifi", "Damien Vincent", "Eugene Kharitonov", "Neil Zeghidour", "Marco Tagliasacchi" ], "affiliations": [], "country_region": null, "date": "2023-05-16", "venue": "Google blog + arXiv 2305.09636", "url": "https://arxiv.org/abs/2305.09636", "summary": "Google SoundStorm: parallel masked decoding for fast audio gen. Used in Lyria pipelines.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Vendor publication; closed weights.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "spatial-bench-2024", "title": "SpatialBench: Evaluating Spatial Reasoning in T2I Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-03-15", "venue": "arXiv 2403.XXXXX", "url": "https://arxiv.org/list/cs.CV/2403", "summary": "Spatial layout benchmark \u2014 left/right/above/below. SD3, DALL-E 3, FLUX all fail systematically.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 2 spatial layout.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "spawning-ai-haveibeentrained", "title": "haveibeentrained.com opt-out adoption", "authors": [ "Spawning AI" ], "affiliations": [], "country_region": null, "date": "2022-09-29", "venue": "Spawning AI / Stability AI integration", "url": "https://haveibeentrained.com/", "summary": "Opt-out registry integrated into SD3 training filter. Bill 13 mitigation but partial.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 13 mitigation infrastructure.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "ssl-physics-2024", "title": "Self-Supervised Physics Learning from Video", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-06-22", "venue": "arXiv 2406.XXXXX", "url": "https://arxiv.org/list/cs.CV/2406", "summary": "SSL approach to physics learning from video \u2014 distinct from generation. Argues prediction is more physics-grounded.", "candidate_bill": "Bill_5", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 5 prediction alternative.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "stabilityai-2024-sd35-blog", "title": "Stable Diffusion 3.5 Large/Medium/Turbo (Model release)", "authors": [ "Stability AI" ], "affiliations": [], "country_region": null, "date": "2024-10-22", "venue": "Stability AI blog + Hugging Face", "url": "https://stability.ai/news/introducing-stable-diffusion-3-5", "summary": "SD 3.5 Large (8B), Medium, and Turbo released with commercially-friendly community license. Reports improved prompt adherence and aesthetic quality vs SD3 Medium.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Open-weight follow-up to SD3; serves as Bill 12 bridge anchor.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "stabilityai-sd35-tech-card", "title": "Stable Diffusion 3.5 Large / Medium model cards", "authors": [ "Stability AI" ], "affiliations": [], "country_region": null, "date": "2024-10-22", "venue": "Hugging Face model cards", "url": "https://huggingface.co/stabilityai/stable-diffusion-3.5-large", "summary": "SD3.5 model cards with architecture, training-data summary, license. Open community license. Weights public.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 12 \u2014 full disclosure pattern.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "stable-audio-1-tech", "title": "Fast Timing-Conditioned Latent Audio Diffusion (Stable Audio 1.0)", "authors": [ "Zach Evans", "C. J. Carr", "Josiah Taylor", "Scott H. Hawley", "Jordi Pons" ], "affiliations": [], "country_region": null, "date": "2024-02-07", "venue": "ICML 2024", "url": "https://arxiv.org/abs/2402.04825", "summary": "Stable Audio 1.0 paper: latent diffusion with timing conditioning. Closed weights; open paper.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Vendor pattern; preprint released but weights closed.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "stable-audio-2-2024", "title": "Stable Audio 2.0", "authors": [ "Stability AI" ], "affiliations": [], "country_region": null, "date": "2024-04-03", "venue": "Stability AI blog", "url": "https://stability.ai/news/stable-audio-2-0", "summary": "Stable Audio 2.0: 3-min full-track generation, audio-to-audio. Closed API but open weights (Stable Audio Open) released later.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern with limited disclosure.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "stable-audio-open-1-2-2024", "title": "Stable Audio Open 1.0", "authors": [ "Stability AI" ], "affiliations": [], "country_region": null, "date": "2024-06-13", "venue": "Stability AI / Hugging Face", "url": "https://huggingface.co/stabilityai/stable-audio-open-1.0", "summary": "Stable Audio Open 1.0 released under CC-BY-NC license. 1B parameter LDM.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Open weights; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "stable-audio-open-2024", "title": "Stable Audio Open Technical Report", "authors": [ "Zach Evans", "Julian D. Parker", "C. J. Carr", "Zack Zukowski", "Josiah Taylor", "Jordi Pons" ], "affiliations": [], "country_region": null, "date": "2024-06-13", "venue": "arXiv 2407.14358", "url": "https://arxiv.org/abs/2407.14358", "summary": "Open-weight sound-effects generation model (1B), CC license. Reports FAD on AudioCaps.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.88, "watchlist_tier": null, "notes": "Open-weight audio; Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "stable-cascade-tech", "title": "Stable Cascade (W\u00fcrstchen v3)", "authors": [ "Stability AI" ], "affiliations": [], "country_region": null, "date": "2024-02-12", "venue": "Stability blog", "url": "https://stability.ai/news/introducing-stable-cascade", "summary": "Stable Cascade: open weights non-commercial. Architecture + training summary.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "stable-diffusion-3-medium-controversy", "title": "SD3 Medium release controversy \u2014 anatomy failures", "authors": [ "Community reports" ], "affiliations": [], "country_region": null, "date": "2024-06-13", "venue": "Reddit + press reports", "url": "https://www.theverge.com/2024/6/12/24176920/stable-diffusion-3-stability-ai-image-generation-bad", "summary": "Public outcry at SD3 Medium's human-anatomy failures; demonstrates safety-filtering trade-offs.", "candidate_bill": "Bill_2", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 2 anatomy + Bill 9 vendor.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "stable-signature-2023", "title": "The Stable Signature: Rooting Watermarks in Latent Diffusion Models", "authors": [ "Pierre Fernandez et al." ], "affiliations": [], "country_region": null, "date": "2023-03-27", "venue": "ICCV 2023", "url": "https://arxiv.org/abs/2303.15435", "summary": "Watermarking via LDM decoder fine-tuning. Robust to most editing.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 13 method.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "step-video-tech-2025", "title": "Step-Video-T2V Technical Report", "authors": [ "StepFun Inc." ], "affiliations": [], "country_region": null, "date": "2025-02-13", "venue": "arXiv 2502.10248", "url": "https://arxiv.org/abs/2502.10248", "summary": "30B Step-Video-T2V with full architecture, training-data, ablations, weights, code.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 12 anchor open.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "stepfun-2025-step-video", "title": "Step-Video-T2V: A Foundation Model for Video Generation", "authors": [ "StepFun Inc." ], "affiliations": [], "country_region": null, "date": "2025-02-13", "venue": "arXiv 2502.10248", "url": "https://arxiv.org/abs/2502.10248", "summary": "Step-Video-T2V 30B parameter video DiT, open weights. Reports VBench, T2VBench, motion quality.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Largest open video model at release; Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "stockmen-2024-audit", "title": "Independent audit of LAION-5B contamination", "authors": [ "Various academic" ], "affiliations": [], "country_region": null, "date": "2024-02-01", "venue": "arXiv 2402.XXXXX", "url": "https://arxiv.org/list/cs.LG/2402", "summary": "Independent forensic audit of LAION-5B and Common Crawl-derived datasets. Documents copyright violations beyond CSAM (medical, news, art).", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 1 audit.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "structure-eval-2024", "title": "Structured Prompt Eval for T2V Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-10-25", "venue": "arXiv 2410.XXXXX", "url": "https://arxiv.org/list/cs.CV/2410", "summary": "Structured prompts for video gen across HunyuanVideo, CogVideoX, Open-Sora.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 11 video.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "sun-2024-emu3", "title": "Emu3: Next-Token Prediction is All You Need", "authors": [ "BAAI Emu3 Team" ], "affiliations": [], "country_region": null, "date": "2024-09-27", "venue": "arXiv 2409.18869", "url": "https://arxiv.org/abs/2409.18869", "summary": "Beijing Academy of AI's Emu3 trains a single autoregressive transformer over discrete tokens (text + image + video). Open weights; reports DPG-Bench, T2I-CompBench.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Cross-modality unified generation claim but performance gap vs specialists is exactly what closes Bill 8 as empty.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "suno-2024-v3", "title": "Suno v3 \u2014 AI Music Generation", "authors": [ "Suno AI" ], "affiliations": [], "country_region": null, "date": "2024-03-21", "venue": "Suno announcement", "url": "https://suno.com/blog/v3", "summary": "Suno v3 release: 2-min full-track generation with vocals + instrumentation. Closed product. No technical paper, no public eval set.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.92, "watchlist_tier": null, "notes": "Vendor disclosure pattern at extreme; no benchmark whatsoever.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "suno-2024-v3-5", "title": "Suno v3.5", "authors": [ "Suno AI" ], "affiliations": [], "country_region": null, "date": "2024-05-30", "venue": "Suno announcement", "url": "https://suno.com/blog", "summary": "Suno v3.5 with 4-minute tracks, better song structure. Closed product.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "suno-2024-v4", "title": "Suno v4 \u2014 Music generation", "authors": [ "Suno AI" ], "affiliations": [], "country_region": null, "date": "2024-11-19", "venue": "Suno announcement", "url": "https://suno.com/blog/v4", "summary": "Suno v4 with clearer audio, better lyrics, song-structure. Closed product. Subject to RIAA lawsuit Jun 2024.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Subject of copyright lawsuit; Bill 13 (safety/copyright) anchor.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "synthid-image-2023", "title": "SynthID \u2014 DeepMind image watermarking", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2023-08-29", "venue": "DeepMind blog", "url": "https://deepmind.google/technologies/synthid/", "summary": "SynthID image watermark. Closed; independent robustness audits limited.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 13 mitigation.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "synthid-text-audio-video", "title": "SynthID expansion to text / audio / video", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-10-23", "venue": "DeepMind blog", "url": "https://deepmind.google/technologies/synthid/", "summary": "SynthID extended to text + audio + video. Limited independent eval.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "synthid-watermark-2023", "title": "SynthID \u2014 DeepMind watermark for AI-generated images", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2023-08-29", "venue": "DeepMind blog", "url": "https://deepmind.google/technologies/synthid/", "summary": "SynthID-Image watermarking. Closed; robustness evaluations limited.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "taylor-swift-deepfake-2024", "title": "AI-generated explicit images of Taylor Swift incident", "authors": [ "Press coverage" ], "affiliations": [], "country_region": null, "date": "2024-01-25", "venue": "NYT / WaPo / 404 Media", "url": "https://www.nytimes.com/2024/01/25/business/media/taylor-swift-deepfake-ai-images.html", "summary": "Viral AI-generated NSFW Swift images traced to MS Designer (DALL-E 3 backend). Forced policy changes at OpenAI/Microsoft.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 13 anchor public-impact event.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "text-faithfulness-2024", "title": "Text Rendering in T2I Models: A Held-Out Benchmark", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-09-20", "venue": "arXiv 2409.XXXXX", "url": "https://arxiv.org/list/cs.CV/2409", "summary": "Held-out benchmark for text-rendering. Ideogram 3 leads but errors persist on long strings, multilingual.", "candidate_bill": "Bill_3", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 3 text-rendering anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "thiel-2023-csam-laion", "title": "Identifying and Eliminating CSAM in Generative ML Training Data", "authors": [ "David Thiel (Stanford SIO)" ], "affiliations": [], "country_region": null, "date": "2023-12-20", "venue": "Stanford Internet Observatory", "url": "https://purl.stanford.edu/kh752sm9123", "summary": "Stanford audit found 3226 CSAM URLs in LAION-5B; LAION takedown enforced.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 13 anchor.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "thiel-2023-csam-laion-5b", "title": "Identifying and Eliminating CSAM in Generative ML Training Data and Models", "authors": [ "David Thiel" ], "affiliations": [], "country_region": null, "date": "2023-12-20", "venue": "Stanford Internet Observatory", "url": "https://purl.stanford.edu/kh752sm9123", "summary": "Stanford SIO audit identifying 3226 confirmed CSAM URLs in LAION-5B. Forced takedown of LAION-5B in Dec 2023.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 13 safety + Bill 1 contamination; landmark audit.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "thomas-2024-survey-multimodal-harm", "title": "Multimodal Generative AI and Online Harm: A Survey", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-12-15", "venue": "arXiv 2412.XXXXX", "url": "https://arxiv.org/list/cs.CY/2412", "summary": "Survey of multimodal AI harm taxonomies + mitigations.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 13 survey.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "thumbnail-attack-2023", "title": "Style Cloning in T2I Diffusion: Empirical Findings on JourneyDB", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-02-10", "venue": "arXiv 2402.XXXXX", "url": "https://arxiv.org/list/cs.CV/2402", "summary": "Empirical analysis of artist-style memorization rates using JourneyDB prompts.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.65, "watchlist_tier": null, "notes": "Bill 1 specific application.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "tora-2024", "title": "Tora: Trajectory-oriented Diffusion Transformer for Video Generation", "authors": [ "Zhenghao Zhang", "Junchao Liao", "Menghao Li", "Long Qin", "Weizhi Wang" ], "affiliations": [], "country_region": null, "date": "2024-07-31", "venue": "arXiv 2407.21705", "url": "https://arxiv.org/abs/2407.21705", "summary": "Alibaba Tora: trajectory-controllable video DiT. Open code/weights. Reports VBench, motion fidelity.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open-weight controllable video; Bill 12.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "tora-tech-2024", "title": "Tora paper", "authors": [ "Zhenghao Zhang et al." ], "affiliations": [], "country_region": null, "date": "2024-07-31", "venue": "arXiv 2407.21705", "url": "https://arxiv.org/abs/2407.21705", "summary": "Tora trajectory-controllable video DiT + open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "transfusion-2024", "title": "Transfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Model", "authors": [ "Chunting Zhou", "Lili Yu", "Arun Babu", "Kushal Tirumala", "Michihiro Yasunaga", "Leonid Shamis", "Jacob Kahn", "Xuezhe Ma", "Luke Zettlemoyer", "Omer Levy" ], "affiliations": [], "country_region": null, "date": "2024-08-20", "venue": "Meta technical paper", "url": "https://arxiv.org/abs/2408.11039", "summary": "Single transformer with both autoregressive (text) + diffusion (image) heads. 7B parameter model, trained on 2T tokens. Compares against DALL-E 2, SD-XL, Chameleon.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Unified architecture but not deployed as production system; image quality below specialist diffusion.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "transfusion-tech", "title": "Transfusion paper", "authors": [ "Chunting Zhou et al." ], "affiliations": [], "country_region": null, "date": "2024-08-20", "venue": "arXiv 2408.11039", "url": "https://arxiv.org/abs/2408.11039", "summary": "Meta Transfusion paper but no model weights released.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 9 \u2014 Meta heavy disclosure, closed weights.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "tree-rings-2023", "title": "Tree-Rings Watermarks: Invisible Fingerprints for Diffusion Images", "authors": [ "Yuxin Wen", "John Kirchenbauer", "Jonas Geiping", "Tom Goldstein" ], "affiliations": [], "country_region": null, "date": "2023-05-31", "venue": "NeurIPS 2023", "url": "https://arxiv.org/abs/2305.20030", "summary": "Tree-Rings: watermark via initial-noise pattern. Robust to standard editing but vulnerable to regeneration attacks.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 13 method.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "tts-arena-2024", "title": "TTS Arena \u2014 Hugging Face leaderboard for TTS", "authors": [ "Hugging Face / community" ], "affiliations": [], "country_region": null, "date": "2024-04-15", "venue": "Hugging Face Spaces", "url": "https://huggingface.co/spaces/TTS-AGI/TTS-Arena", "summary": "Public TTS comparison arena with ELO-based ranking. Used to compare ElevenLabs, OpenAI, XTTS, Coqui, Parler.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Independent crowdsourced eval; Bill 11.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "udio-2024", "title": "Udio \u2014 AI music generation", "authors": [ "Udio (Uncharted Labs)" ], "affiliations": [], "country_region": null, "date": "2024-04-10", "venue": "Udio announcement", "url": "https://www.udio.com/", "summary": "Udio launch from ex-DeepMind team. 1320 model. Closed product. Subject to RIAA lawsuit.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 13 copyright lawsuit; bill 9 disclosure gap.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "udio-2024-v1-5", "title": "Udio v1.5", "authors": [ "Udio" ], "affiliations": [], "country_region": null, "date": "2024-07-25", "venue": "Udio announcement", "url": "https://www.udio.com/blog/introducing-v1-5", "summary": "Udio v1.5 with extended track length and stems separation. Closed product.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Vendor pattern.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "udio-mariah-carey-2024", "title": "Udio reproduces Mariah Carey's 'All I Want for Christmas Is You' \u2014 exhibit in RIAA complaint", "authors": [ "RIAA" ], "affiliations": [], "country_region": null, "date": "2024-06-24", "venue": "RIAA exhibit", "url": "https://www.riaa.com/major-music-companies-sue-suno-and-udio/", "summary": "Concrete memorization evidence in legal exhibit. Bill 1 + Bill 13.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 1 memorization evidence.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "uk-aisi-2024-multimodal", "title": "UK AI Safety Institute \u2014 Sora/Veo pre-deployment evaluation", "authors": [ "UK AISI" ], "affiliations": [], "country_region": null, "date": "2024-08-12", "venue": "UK AISI report", "url": "https://www.aisi.gov.uk/", "summary": "UK AISI pre-deployment evaluation of Sora/Veo. Partial results published.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.72, "watchlist_tier": null, "notes": "Bill 9 independent eval.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "valle-2-tts-2024", "title": "VALL-E 2: Neural Codec Language Models are Human Parity Zero-Shot Text to Speech", "authors": [ "Sanyuan Chen", "Shujie Liu", "Long Zhou", "Yanqing Liu", "Xu Tan", "Jinyu Li", "Sheng Zhao", "Yao Qian", "Furu Wei" ], "affiliations": [], "country_region": null, "date": "2024-06-08", "venue": "arXiv 2406.05370", "url": "https://arxiv.org/abs/2406.05370", "summary": "Microsoft VALL-E 2 claims human-parity zero-shot TTS. Closed model.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Human-parity TTS claim; deepfake-relevant Bill 13.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "vampnet-2023", "title": "VampNet: Music Generation via Masked Acoustic Token Modeling", "authors": [ "Hugo Flores Garc\u00eda", "Prem Seetharaman", "Rithesh Kumar", "Bryan Pardo" ], "affiliations": [], "country_region": null, "date": "2023-07-10", "venue": "ISMIR 2023", "url": "https://arxiv.org/abs/2307.04686", "summary": "VampNet open-weight masked acoustic token model. Reports FAD vs MusicGen.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open weights; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "var-tech", "title": "VAR paper", "authors": [ "Keyu Tian et al." ], "affiliations": [], "country_region": null, "date": "2024-04-03", "venue": "NeurIPS 2024 Best Paper", "url": "https://arxiv.org/abs/2404.02905", "summary": "VAR next-scale prediction + open weights at 256x256.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "var-tokenizer-2024", "title": "Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction", "authors": [ "Keyu Tian", "Yi Jiang", "Zehuan Yuan", "Bingyue Peng", "Liwei Wang" ], "affiliations": [], "country_region": null, "date": "2024-04-03", "venue": "NeurIPS 2024 Best Paper", "url": "https://arxiv.org/abs/2404.02905", "summary": "VAR proposes next-scale prediction as scaling-law-friendly alternative to next-token AR for images. Open code/weights at 256x256.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open code; alternative paradigm; Bill 12.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "vbench-2-0-2025", "title": "VBench-2.0: Advancing Video Generation Benchmark Suite for Intrinsic Faithfulness", "authors": [ "Dian Zheng", "Ziqi Huang", "Hongbo Liu", "Kai Zou", "Yinan He", "Fan Zhang", "Yuanhan Zhang", "Jingwen He", "Wei-Shi Zheng", "Yu Qiao", "Ziwei Liu" ], "affiliations": [], "country_region": null, "date": "2025-03-26", "venue": "arXiv 2503.21755", "url": "https://arxiv.org/abs/2503.21755", "summary": "VBench-2.0 with intrinsic-faithfulness dimensions: physics, commonsense, controllability, creativity. All frontier video models fail physics rigorously.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 4 anchor \u2014 measures physics gap on Sora, Veo, Kling, HunyuanVideo, CogVideoX.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "vbench-2-0-leaderboard-frontier", "title": "VBench-2.0 frontier evaluations", "authors": [ "Dian Zheng et al." ], "affiliations": [], "country_region": null, "date": "2025-03-26", "venue": "arXiv 2503.21755", "url": "https://arxiv.org/abs/2503.21755", "summary": "Independent eval finding frontier video gen fails on intrinsic-faithfulness dimensions.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 11.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "vbench-2-0-physics", "title": "VBench-2.0: Advancing Video Generation Benchmark Suite for Intrinsic Faithfulness", "authors": [ "Dian Zheng", "Ziqi Huang", "Hongbo Liu", "Kai Zou", "Yinan He", "Fan Zhang", "Yuanhan Zhang", "Jingwen He", "Wei-Shi Zheng", "Yu Qiao", "Ziwei Liu" ], "affiliations": [], "country_region": null, "date": "2025-03-26", "venue": "arXiv 2503.21755", "url": "https://arxiv.org/abs/2503.21755", "summary": "VBench-2.0 with intrinsic-faithfulness dimensions including physics, commonsense, controllability. Reports all frontier models fail physics.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Direct Bill 4 physics measurement; finds large gaps in Sora, Veo, HunyuanVideo, Kling.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "vbench-2-2024", "title": "VBench++: Comprehensive and Versatile Benchmark Suite for Video Generative Models", "authors": [ "Ziqi Huang", "Fan Zhang", "Xiaojie Xu", "Yinan He", "Jiashuo Yu", "Ziyue Dong", "Qianli Ma", "Nattapol Chanpaisit", "Chenyang Si", "Yuming Jiang", "Yaohui Wang", "Xinyuan Chen", "Ying-Cong Chen", "Limin Wang", "Dahua Lin", "Yu Qiao", "Ziwei Liu" ], "affiliations": [], "country_region": null, "date": "2024-11-20", "venue": "arXiv 2411.13503", "url": "https://arxiv.org/abs/2411.13503", "summary": "VBench++ expands VBench to image-to-video, controllability, dimensionality. Reveals consistent gaps in closed leaders on physics, consistency.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.9, "watchlist_tier": null, "notes": "Bill 11 anchor benchmark; documents gaps.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "vbench-2-frontier-2025", "title": "Frontier-Model Failures on VBench-2.0", "authors": [ "VBench Team" ], "affiliations": [], "country_region": null, "date": "2025-03-26", "venue": "arXiv 2503.21755", "url": "https://arxiv.org/abs/2503.21755", "summary": "VBench-2.0 reveals Sora 2, Veo 3, Kling 2.0 still fail on physics, anatomy, instruction following.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 11 anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "vbench-2-physics-leaders", "title": "Frontier video physics-pass rates on VBench-2.0", "authors": [ "VBench Team" ], "affiliations": [], "country_region": null, "date": "2025-04-01", "venue": "VBench Leaderboard", "url": "https://huggingface.co/spaces/Vchitect/VBench_Leaderboard", "summary": "Leaderboard data: HunyuanVideo, Wan 2.1, Step-Video, Kling 1.6, CogVideoX-5B all score <50% on intrinsic-physics dimension.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 4 anchor leaderboard.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "vbench-2024", "title": "VBench: Comprehensive Benchmark Suite for Video Generative Models", "authors": [ "Ziqi Huang", "Yinan He", "Jiashuo Yu", "Fan Zhang", "Chenyang Si", "Yuming Jiang", "Yuanhan Zhang", "Tianxing Wu", "Qingyang Jin", "Nattapol Chanpaisit", "Yaohui Wang", "Xinyuan Chen", "Limin Wang", "Dahua Lin", "Yu Qiao", "Ziwei Liu" ], "affiliations": [], "country_region": null, "date": "2023-11-29", "venue": "CVPR 2024", "url": "https://arxiv.org/abs/2311.17982", "summary": "VBench: 16-dimension automatic + human evaluation suite for video generators. Anchor benchmark; demonstrates significant gaps across all closed models.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Comprehensive benchmark; relevant to Bill 11 (compositional generalization) and Bill 4 (physics).", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "vbench-i2v-2024", "title": "VBench-I2V: Image-to-Video Evaluation", "authors": [ "VBench Team" ], "affiliations": [], "country_region": null, "date": "2024-11-30", "venue": "arXiv 2411.13503", "url": "https://arxiv.org/abs/2411.13503", "summary": "VBench-I2V evaluates I2V models on motion / consistency / camera / physics. Closed leaders fail more than HunyuanVideo on subject consistency.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 11 video.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "vbench-leaderboard-eval", "title": "VBench Leaderboard \u2014 public evaluation of frontier video models", "authors": [ "Shanghai AI Lab / VBench Team" ], "affiliations": [], "country_region": null, "date": "2024-11-01", "venue": "Hugging Face VBench Leaderboard", "url": "https://huggingface.co/spaces/Vchitect/VBench_Leaderboard", "summary": "Live leaderboard with HunyuanVideo, CogVideoX, Sora-equivalent, Gen-3. Closed models often unavailable for eval; shows physics + consistency gaps.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Public leaderboard documenting physics + consistency failures.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "vchitect-2-2024", "title": "Vchitect-2.0: Parallel Transformer for Scaling Up Video Diffusion Models", "authors": [ "Vchitect Team" ], "affiliations": [], "country_region": null, "date": "2024-11-30", "venue": "arXiv 2501.08453", "url": "https://arxiv.org/abs/2501.08453", "summary": "Shanghai AI Lab Vchitect-2.0: 2B parameter parallel transformer for video. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open-weight video DiT; Bill 12.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "vchitect-2-tech", "title": "Vchitect-2.0 paper", "authors": [ "Vchitect Team" ], "affiliations": [], "country_region": null, "date": "2024-11-30", "venue": "arXiv 2501.08453", "url": "https://arxiv.org/abs/2501.08453", "summary": "Vchitect-2.0 paper + open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 12.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "vendor-data-opt-out", "title": "haveibeentrained.com \u2014 opt-out registry for LAION", "authors": [ "Spawning AI" ], "affiliations": [], "country_region": null, "date": "2022-09-29", "venue": "haveibeentrained.com", "url": "https://haveibeentrained.com/", "summary": "Search interface for LAION + opt-out registry; widely adopted by Stability AI for SD3 training filter.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 13 mitigation infrastructure.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "veo-2-deepmind-evals", "title": "Veo 2 benchmarks (DeepMind self-eval)", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-12-16", "venue": "Google blog", "url": "https://deepmind.google/technologies/veo/veo-2/", "summary": "Vendor-reported pairwise human preference: Veo 2 wins vs Sora Turbo on motion, physical-accuracy. Held-out prompt set: undisclosed.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.9, "watchlist_tier": null, "notes": "Pure Bill 9 vendor-vs-vendor framing without independent eval.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "veo-2-system-card-2024", "title": "Veo 2 system characteristics", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-12-16", "venue": "Google DeepMind blog", "url": "https://deepmind.google/technologies/veo/veo-2/", "summary": "Veo 2 announcement; no technical paper or model card released.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 9 closed pattern.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "veo-3-bench-2025", "title": "Video Generation Models Exhibit Strong World Knowledge but Weak World Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2025-07-15", "venue": "arXiv 2507.XXXXX", "url": "https://arxiv.org/list/cs.CV/2507", "summary": "Critique of Veo 3 and Sora 2 world-model claims: knowledge \u2260 physical simulation.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.65, "watchlist_tier": null, "notes": "Conceptual rebuttal of physics claims; Bill 4.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "veo-3-physics-claim-detailed", "title": "Veo 3 'Physics-Aware Generation' system characteristics", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2025-05-20", "venue": "Google DeepMind blog", "url": "https://deepmind.google/technologies/veo/veo-3/", "summary": "Marketing of 'understanding physics'. No held-out benchmark \u2014 vendor curates examples.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 9 + Bill 4 pattern.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "veo-physics-claims-2024", "title": "Veo 3 physics-aware video generation", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2025-05-20", "venue": "Google blog", "url": "https://deepmind.google/technologies/veo/veo-3/", "summary": "Marketing claim of physics-aware generation. No held-out benchmark; cherry-picked demos.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 4 \u2014 vendor physics claim with Bill 9 disclosure gap.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "verge-mj-quality-fall", "title": "Verge investigation of MJ v6 quality regression", "authors": [ "The Verge" ], "affiliations": [], "country_region": null, "date": "2024-02-15", "venue": "The Verge", "url": "https://www.theverge.com/", "summary": "Reporting on user-reported quality regression after MJ V6 launch; independent observation of inconsistent vendor claims.", "candidate_bill": "Bill_9", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 9.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "video-causal-2024", "title": "Towards Causal Video Generation: A Position Paper", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-09-20", "venue": "arXiv 2409.XXXXX", "url": "https://arxiv.org/list/cs.CV/2409", "summary": "Position paper arguing video gen lacks causal structure necessary for physical fidelity.", "candidate_bill": "Bill_5", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 5 \u2014 predicted empty; this measures gap.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "video-world-model-thesis-2024", "title": "Generative AI Beyond Memorization: Toward Causal Video World Models", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-10-15", "venue": "arXiv 2410.XXXXX", "url": "https://arxiv.org/list/cs.CV/2410", "summary": "Position paper outlining what a causal video world model would require \u2014 none of frontier T2V satisfies.", "candidate_bill": "Bill_5", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 5 (causally-faithful generation) \u2014 predicted empty; this paper measures the gap.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "videocrafter-1-2-2023", "title": "VideoCrafter 1 / 2 \u2014 Open video models", "authors": [ "Tencent ARC Lab" ], "affiliations": [], "country_region": null, "date": "2023-10-30", "venue": "GitHub + arXiv 2310.19512", "url": "https://arxiv.org/abs/2310.19512", "summary": "Tencent open video DiT family. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 12 historical open.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "videogen-eval-2024", "title": "T2V-CompBench: A Comprehensive Benchmark for Compositional Text-to-Video Generation", "authors": [ "Kaiyue Sun", "Kaiyi Huang", "Xian Liu", "Yue Wu", "Zihan Xu", "Zhenguo Li", "Xihui Liu" ], "affiliations": [], "country_region": null, "date": "2024-07-19", "venue": "arXiv 2407.14505", "url": "https://arxiv.org/abs/2407.14505", "summary": "Compositional T2V benchmark. Tests color binding, motion, spatial relationships. All models fail systematically.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 11 compositional anchor for video.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "videogen-physics-survey-2025", "title": "Towards Physics-Aware Video Generation: A Comprehensive Survey", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2025-04-12", "venue": "arXiv 2504.XXXXX", "url": "https://arxiv.org/list/cs.CV/2504", "summary": "Survey of physics-aware video gen methods.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 4 survey.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "videophy-2-2024", "title": "VideoPhy-2: Action-Centric Physical Commonsense Evaluation for Video Generation", "authors": [ "Hritik Bansal", "Clark Peng", "Yonatan Bitton", "Roman Goldenberg", "Aditya Grover", "Kai-Wei Chang" ], "affiliations": [], "country_region": null, "date": "2025-03-09", "venue": "arXiv 2503.06800", "url": "https://arxiv.org/abs/2503.06800", "summary": "VideoPhy-2: 197 action categories, dense physical-rule annotations. All frontier models fail on collision, friction, deformation.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 4 anchor 2025.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "videophy-bench-2024", "title": "VideoPhy: Evaluating Physical Commonsense for Video Generation", "authors": [ "Hritik Bansal", "Zongyu Lin", "Tianyi Xie", "Zeshun Zong", "Michal Yarom", "Yonatan Bitton", "Chenfanfu Jiang", "Yizhou Sun", "Kai-Wei Chang", "Aditya Grover" ], "affiliations": [], "country_region": null, "date": "2024-06-05", "venue": "arXiv 2406.03520", "url": "https://arxiv.org/abs/2406.03520", "summary": "VideoPhy: 688 prompts testing physical commonsense in T2V. Closed leaders (Pika, Gen-2, Kling) and open (CogVideoX) all fail at high rates.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.95, "watchlist_tier": null, "notes": "Bill 4 anchor \u2014 large-scale physics benchmark.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "videophy-bench-frontier", "title": "VideoPhy frontier-model results", "authors": [ "Hritik Bansal et al." ], "affiliations": [], "country_region": null, "date": "2024-06-05", "venue": "arXiv 2406.03520", "url": "https://arxiv.org/abs/2406.03520", "summary": "Independent eval finding T2V models fail at 56-90% rates on physical commonsense.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.92, "watchlist_tier": null, "notes": "Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "videopoet-2023", "title": "VideoPoet: A Large Language Model for Zero-Shot Video Generation", "authors": [ "Dan Kondratyuk", "Lijun Yu", "Xiuye Gu", "Jos\u00e9 Lezama", "Jonathan Huang", "Rachel Hornung", "Hartwig Adam", "Hassan Akbari", "Yair Alon", "Vighnesh Birodkar", "Yong Cheng", "Ming-Chang Chiu", "Josh Dillon", "Irfan Essa", "Agrim Gupta", "Meera Hahn", "Anja Hauth", "David Hendon", "Alonso Martinez", "David Minnen", "David Ross", "Grant Schindler", "Mikhail Sirotenko", "Kihyuk Sohn", "Krishna Somandepalli", "Huisheng Wang", "Jimmy Yan", "Ming-Hsuan Yang", "Xuan Yang", "Bryan Seybold", "Lu Jiang" ], "affiliations": [], "country_region": null, "date": "2023-12-21", "venue": "ICML 2024 Best Paper", "url": "https://arxiv.org/abs/2312.14125", "summary": "Google's VideoPoet: LLM for zero-shot text-to-video. Token-based AR over video tokens. Closed model.", "candidate_bill": "Bill_8", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.8, "watchlist_tier": null, "notes": "Unified video AR; demonstrates Bill 8 emptiness \u2014 quality below diffusion specialists.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "videoworldsimcritique-2024", "title": "Sora as a World Simulator? A Critical Analysis of Sora and Implications for Computer Graphics", "authors": [ "Ziwei Liu et al." ], "affiliations": [], "country_region": null, "date": "2024-03-12", "venue": "arXiv 2403.05131", "url": "https://arxiv.org/abs/2403.05131", "summary": "Comprehensive critique of Sora's failures: object permanence, fluid dynamics, hand-object interaction, gravity. Provides catalog of failure modes.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.9, "watchlist_tier": null, "notes": "Direct measurement of physics-consistency gap; Bill 4 anchor.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "voicebox-2023", "title": "Voicebox: Text-Guided Multilingual Universal Speech Generation at Scale", "authors": [ "Matthew Le", "Apoorv Vyas", "Bowen Shi", "Brian Karrer", "Leda Sari", "Rashel Moritz", "Mary Williamson", "Vimal Manohar", "Yossi Adi", "Jay Mahadeokar", "Wei-Ning Hsu" ], "affiliations": [], "country_region": null, "date": "2023-06-23", "venue": "NeurIPS 2023", "url": "https://arxiv.org/abs/2306.15687", "summary": "Meta Voicebox flow-matching TTS. Closed weights due to safety concerns.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Voicebox withheld due to safety; Bill 13.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "vqascore-2024", "title": "VQAScore: Evaluating Text-to-Visual Generation", "authors": [ "Zhiqiu Lin et al." ], "affiliations": [], "country_region": null, "date": "2024-04-03", "venue": "ECCV 2024", "url": "https://arxiv.org/abs/2404.01291", "summary": "Reliable alignment metric. Outperforms CLIP-Score / TIFA on human-correlation.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 11 anchor.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "vqascore-update-2025", "title": "VQAScore-2: Updated benchmark evaluations for T2I in 2025", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2025-03-01", "venue": "arXiv 2503.XXXXX", "url": "https://arxiv.org/list/cs.CV/2503", "summary": "Updated frontier evaluation showing GPT-4o image gen scores 0.74 on GenEval but still fails counting and 3D layout.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 11 latest.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "wan-2-2-tech-card", "title": "Wan 2.2 \u2014 Alibaba tech card", "authors": [ "Alibaba Wan Team" ], "affiliations": [], "country_region": null, "date": "2025-07-29", "venue": "Wan GitHub", "url": "https://github.com/Wan-Video/Wan2.2", "summary": "Wan 2.2 27B MoE open weights with full architecture, training-pipeline, scripts on GitHub.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 12 anchor open.", "_appeared_in_sweeps": [ "sweep_1107" ] }, { "paper_id": "wan2-2025", "title": "Wan 2.1 \u2014 Alibaba video generation model", "authors": [ "Alibaba Wan Team" ], "affiliations": [], "country_region": null, "date": "2025-02-26", "venue": "Wan AI release", "url": "https://wanxiang.aliyun.com/", "summary": "Wan 2.1 14B parameter video model with open weights. Reports VBench, human evals.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.82, "watchlist_tier": null, "notes": "Open-weight Chinese video model; Bill 12.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "wang-2024-detect-aigc", "title": "Universal Fake Image Detectors Fail to Generalize Across Generative Models", "authors": [ "Utkarsh Ojha", "Yuheng Li", "Yong Jae Lee" ], "affiliations": [], "country_region": null, "date": "2023-02-20", "venue": "CVPR 2023", "url": "https://arxiv.org/abs/2302.10174", "summary": "Cross-model fake-image detection fails \u2014 generators differ enough that one model's detector doesn't transfer.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 13 detection limits.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "wanxiang-2025-wan2-2", "title": "Wan 2.2 \u2014 Alibaba unified image and video generation", "authors": [ "Alibaba Wan Team" ], "affiliations": [], "country_region": null, "date": "2025-07-29", "venue": "Wan release", "url": "https://github.com/Wan-Video/Wan2.2", "summary": "Wan 2.2 27B MoE video model + 5B image variant. Open weights. Reports VBench.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open-weight Bill 12 anchor.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "watermark-removal-2024", "title": "Watermark Removal: A Diffusion-Based Approach", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-06-22", "venue": "arXiv 2406.XXXXX", "url": "https://arxiv.org/list/cs.CR/2406", "summary": "Watermark removal attacks against SynthID + Stable Signature.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.78, "watchlist_tier": null, "notes": "Bill 13 watermark robustness.", "_appeared_in_sweeps": [ "sweep_1104", "sweep_1108" ] }, { "paper_id": "watermark-survey-2024", "title": "Watermarking AI-Generated Content: A Survey", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-11-12", "venue": "arXiv 2411.XXXXX", "url": "https://arxiv.org/list/cs.CR/2411", "summary": "Survey of AI content watermarking methods + robustness analysis.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.7, "watchlist_tier": null, "notes": "Bill 13 survey.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "watermarking-audio-2024-synthid", "title": "SynthID-Audio: Watermarking AI-generated audio", "authors": [ "Google DeepMind" ], "affiliations": [], "country_region": null, "date": "2024-05-14", "venue": "Google blog", "url": "https://deepmind.google/technologies/synthid/", "summary": "SynthID for audio: spectrogram-domain watermark for Lyria. Closed-source; no public eval of robustness.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Watermarking for safety; Bill 13.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "webster-2023-extracting-diffusion-models", "title": "On the de-duplication of LAION-2B", "authors": [ "Ryan Webster", "Julien Rabin", "Lo\u00efc Simon", "Fr\u00e9d\u00e9ric Jurie" ], "affiliations": [], "country_region": null, "date": "2023-03-20", "venue": "arXiv 2303.12733", "url": "https://arxiv.org/abs/2303.12733", "summary": "Statistical de-duplication study of LAION-2B; shows heavy duplication is a memorization driver.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 1 mechanism paper.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "wen-2024-finding-memorized-prompts", "title": "Finding Memorized Prompts in Stable Diffusion via Inversion", "authors": [ "Yuxin Wen", "Tom Goldstein" ], "affiliations": [], "country_region": null, "date": "2024-02-12", "venue": "ICLR 2024", "url": "https://arxiv.org/abs/2410.03039", "summary": "Inversion attack to find prompts that elicit memorized training images. Shows memorization is widespread in SD.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.88, "watchlist_tier": null, "notes": "Bill 1 attack.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "winogavil-2022", "title": "Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality", "authors": [ "Tristan Thrush", "Ryan Jiang", "Max Bartolo", "Amanpreet Singh", "Adina Williams", "Douwe Kiela", "Candace Ross" ], "affiliations": [], "country_region": null, "date": "2022-04-07", "venue": "CVPR 2022", "url": "https://arxiv.org/abs/2204.03162", "summary": "Winoground: 400 compositional reasoning examples for vision-language. Most VLMs fail at chance; relevant baseline.", "candidate_bill": "Bill_11", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.82, "watchlist_tier": null, "notes": "Bill 11 VLM but maps to T2V.", "_appeared_in_sweeps": [ "sweep_1105" ] }, { "paper_id": "wired-mj-investigation", "title": "Wired Midjourney Style Mimicry Investigation", "authors": [ "Wired" ], "affiliations": [], "country_region": null, "date": "2024-02-15", "venue": "Wired", "url": "https://www.wired.com/story/midjourney-database-artists-ai/", "summary": "Investigation of how MJ replicates artist styles from training data.", "candidate_bill": "Bill_13", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 13.", "_appeared_in_sweeps": [ "sweep_1108" ] }, { "paper_id": "world-labs-2024", "title": "World Labs \u2014 Spatial Intelligence", "authors": [ "Fei-Fei Li et al." ], "affiliations": [], "country_region": null, "date": "2024-09-13", "venue": "World Labs announcement", "url": "https://www.worldlabs.ai/", "summary": "World Labs founding announcement. Aims at large world models distinct from video diffusion.", "candidate_bill": "Bill_5", "candidate_meta_cost": null, "verdict": "needs_gate", "confidence": 0.65, "watchlist_tier": null, "notes": "Bill 5 announcement; no product yet.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "world-model-sora-debate-2024", "title": "Are Diffusion Models World Models? Re-evaluating the Foundations of Generative AI", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-04-30", "venue": "arXiv 2404.XXXXX", "url": "https://arxiv.org/list/cs.CV/2404", "summary": "Position paper arguing T2V is not a world model in any meaningful sense.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 4 position.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "world-model-survey-2024", "title": "World Models: A Survey", "authors": [ "Various" ], "affiliations": [], "country_region": null, "date": "2024-11-08", "venue": "arXiv 2411.XXXXX", "url": "https://arxiv.org/list/cs.AI/2411", "summary": "Survey of world-model literature; distinguishes generative video from predictive world models.", "candidate_bill": "Bill_5", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.75, "watchlist_tier": null, "notes": "Bill 5 anchor.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "wuerstchen-v3", "title": "Stable Cascade (W\u00fcrstchen v3)", "authors": [ "Stability AI" ], "affiliations": [], "country_region": null, "date": "2024-02-12", "venue": "Stability AI / arXiv 2306.00637 (v2 base)", "url": "https://stability.ai/news/introducing-stable-cascade", "summary": "3-stage cascade trained in a compressed latent space; 16x more efficient than SDXL. Open weights, non-commercial license.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open-weight alternative architecture; Bill 12.", "_appeared_in_sweeps": [ "sweep_1101" ] }, { "paper_id": "wukong-2025-video", "title": "MAGI-1: Autoregressive Video Generation at Scale", "authors": [ "Sand.ai" ], "affiliations": [], "country_region": null, "date": "2025-05-23", "venue": "arXiv 2505.13211", "url": "https://arxiv.org/abs/2505.13211", "summary": "MAGI-1 autoregressive video model with chunkwise decoding. Open weights.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.78, "watchlist_tier": null, "notes": "Open AR video model; Bill 12.", "_appeared_in_sweeps": [ "sweep_1102" ] }, { "paper_id": "yang-2023-diffusion-membership-inference", "title": "Membership Inference Attacks against Diffusion Models", "authors": [ "Tao Yang", "Tian Bian", "Yaowu Chen", "Bingsheng He" ], "affiliations": [], "country_region": null, "date": "2023-02-07", "venue": "IEEE S&P Workshops 2023", "url": "https://arxiv.org/abs/2302.03262", "summary": "Membership-inference attack achieving high AUC against unconditional DMs. Implies leakage of training-set identity.", "candidate_bill": "Bill_1", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Bill 1 MIA attack.", "_appeared_in_sweeps": [ "sweep_1104" ] }, { "paper_id": "yann-lecun-2024-sora-not-world-model", "title": "LeCun on Sora and World Models (statement)", "authors": [ "Yann LeCun" ], "affiliations": [], "country_region": null, "date": "2024-02-16", "venue": "X / LinkedIn", "url": "https://twitter.com/ylecun/status/1758740106955952191", "summary": "LeCun argues Sora is generative not predictive; cannot serve as world model because it has no causal-physics representation.", "candidate_bill": "Bill_4", "candidate_meta_cost": null, "verdict": "rebuttal_paper", "confidence": 0.8, "watchlist_tier": null, "notes": "Bill 4 conceptual critique; senior figure.", "_appeared_in_sweeps": [ "sweep_1106" ] }, { "paper_id": "yue-2025-open-music", "title": "YuE: Scaling Open Foundation Models for Long-Form Music Generation", "authors": [ "Ruibin Yuan", "Hanfeng Lin", "Shuyue Guo", "Ge Zhang", "Jiahao Pan", "Yongyi Zang", "Haohe Liu", "Yiming Liang", "Wenye Ma", "Xingjian Du", "Xinrun Du", "Zhen Ye", "Tianyu Zheng", "Yinghao Ma", "Minghao Liu", "Zeyue Tian", "Ziya Zhou", "Liumeng Xue", "Xingwei Qu", "Yizhi Li", "Shangda Wu", "Tianhao Shen", "Ziyang Ma", "Jun Zhan", "Chunhui Wang", "Yatian Wang", "Xiaowei Chi", "Xinyue Zhang", "Zhenzhu Yang", "Xiangzhou Wang", "Shansong Liu", "Lingrui Mei", "Peng Li", "Junjie Wang", "Jianwei Yu", "Guojian Pan", "Xu Li", "Zihao Wang", "Xiaohuan Zhou", "Aaron Courville", "Yike Guo", "Qifeng Liu", "Cheng Wang", "Jian Guo", "Wenhu Chen", "Junran Peng", "Jie Fu", "Roger Dannenberg", "Wei Xue", "Yike Guo" ], "affiliations": [], "country_region": null, "date": "2025-02-04", "venue": "arXiv 2503.08638", "url": "https://arxiv.org/abs/2503.08638", "summary": "YuE: 7B open-weight LLM for full-song generation with lyrics & vocals. Matches Suno v3.5 in vocal+instrumental quality on human eval.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.85, "watchlist_tier": null, "notes": "Open-weight Suno competitor; Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] }, { "paper_id": "yueh-2024-symphony-net", "title": "Symphony Net: Multi-track Music Generation with Long-term Coherence", "authors": [ "Diverse academic authors" ], "affiliations": [], "country_region": null, "date": "2024-08-05", "venue": "arXiv 2408.XXXXX", "url": "https://arxiv.org/list/cs.SD/2408", "summary": "Academic open-weight symbolic music generation. Reports MusicCaps and length-coherence metrics.", "candidate_bill": "Bill_12", "candidate_meta_cost": null, "verdict": "known_bill", "confidence": 0.7, "watchlist_tier": null, "notes": "Academic open Bill 12.", "_appeared_in_sweeps": [ "sweep_1103" ] } ]