Each generated condition passed through BuffData strict mode while its config requested all PII redaction, minhash deduplication, LLM filtering, and forced reclassification. Strict mode overrode every mutating stage. Original and generated conditions use the same AG News test set, vocabulary, model, epochs, and seeds.
| Train rows | Condition | Accuracy | Macro-F1 | Text/label identical |
|---|---|---|---|---|
| 20,000 | original | 0.8811 ± 0.0016 | 0.8811 ± 0.0016 | yes |
| 20,000 | strict_generated | 0.8811 ± 0.0016 | 0.8811 ± 0.0016 | yes |
| 100,000 | original | 0.8964 ± 0.0011 | 0.8962 ± 0.0011 | yes | | 100,000 | strict_generated | 0.8964 ± 0.0011 | 0.8962 ± 0.0011 | yes |
- 20k strict parity: PASS
- 100k strict parity: PASS
A passing gate requires the complete valid labeled row sequence, every classification text, every label, accuracy, and macro-F1 to match the original exactly.