Claim-level factuality verification on captured-agent traces (held-out)
0.673Block PrecisionProvenanceGuard
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ProvenanceGuardClaims=3612026.06 | 0.673 | 0.993 | 0.802 | 0.812 | 0.406 | |
| RAGAS FaithfulnessClaims=3612026.06 | 0.66 | 0.892 | 0.758 | 0.781 | 0.39 | |
| MiniCheckClaims=3612026.06 | 0.655 | 0.971 | 0.783 | 0.792 | 0.396 | |
| AlignScoreClaims=3612026.06 | 0.51 | 0.942 | 0.662 | 0.629 | 0.313 | |
| SummaC-ZSClaims=3612026.06 | 0.433 | 0.439 | 0.436 | 0.562 | 0.27 |