# Factorization Data Schema

## Corpus Synthesis JSON

`corpus_synthesis.json` is a synthesis export with these top-level fields:

- `scan_date`
- `scan_round`
- `classifier_version`
- `method`
- `benchmark`
- `atlas_state`
- `papers`

## Heatmap CSV

`heatmap.csv` uses:

```csv
bill,candidates,clean_triggers,status,notes
```

`clean_triggers=0` on starred rows means the candidate papers either pay a meta-cost, are rebutted in the corpus, or fall through an escape gate.

## Source Card Manifest

`source_card_manifest.json` groups rows as:

- `verified_public`
- `quarantined`
- `internal_or_synthesis`

Only `verified_public` rows should be treated as public-source-carded evidence.