# Screening Benchmark v0 result files

- `metrics.json`: aggregate and per-corpus results plus display samples.
- `predictions.csv`: every record, gold label, LitSynth decision, confidence, rationale, and evidence quotes.
- `false-excludes.csv`: every gold-relevant record that LitSynth auto-excluded.
- `manifest.json`: frozen corpus sources, labels, protocols, hashes, and license.
- `methodology.md`: evaluation method and limitations.
- `run.json`: model and execution metadata.