Faithfulness Detection on FEVER n=200
1M4HHEM-2.1-Open
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HHEM-2.1-OpenMode=in-dist on FEVER, Four-way partition=no, Tiered decision=no2026.04 | 1 | 86 | |
| RAGAS faithfulnessJudge=gpt-5.4, Four-way partition=no, Tiered decision=no2026.04 | 0.63 | 88 | |
| GSARMode=adversarial mode, Judge=Opus 4.7, Four-way partition=yes, Tiered decision=yes2026.04 | 0.31 | — | |
| GSARMode=adversarial mode, Judge=gpt-5.4, Four-way partition=yes, Tiered decision=yes2026.04 | 0.27 | — | |
| GSARMode=single mode, Judge=gpt-5.4, Four-way partition=yes, Tiered decision=yes2026.04 | 0.24 | — | |
| GSARMode=single mode, Judge=Opus 4.7, Four-way partition=yes, Tiered decision=yes2026.04 | 0.06 | — |