Discriminative Hallucination Evaluation on AMBER (Acc, P, R, F1)
84.3AccuracyMESA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MESA2026.04 | 84.3 | 90.3 | 85.4 | 87.8 | |
| Nullu2026.04 | 81.5 | 87.4 | 84.2 | 85.8 | |
| PSRDBase Model=LLaVA-1.5-7B2026.04 | 81.2 | — | — | 85 | |
| ICT2026.04 | 79.5 | 84.7 | 84.4 | 84.5 | |
| VTI2026.04 | 79.1 | 82.1 | 87.6 | 84.8 | |
| OctopusBase Model=LLaVA-1.5-7B2026.04 | 76.7 | — | — | 82.7 | |
| ICD2026.04 | 74.1 | 73.2 | 96.2 | 83.1 | |
| VCD2026.04 | 73.9 | 83.8 | 75.2 | 79.3 | |
| Shikra-7BVariant=Original2026.04 | 73.5 | 85.7 | 72.8 | 78.7 | |
| Shikra-7BVariant=Ours2026.04 | 73.3 | 85.1 | 74.2 | 79.3 | |
| Vanilla2026.04 | 71.4 | 81.9 | 73.1 | 77.2 | |
| LLaVA-1.5-7BVariant=Ours2026.04 | 71.1 | 89.5 | 63.9 | 74.6 | |
| AVISCBase Model=LLaVA-1.5-7B2026.04 | 70.7 | — | — | 75.5 | |
| LLaVA-1.5-7BVariant=Original2026.04 | 70 | 90 | 61.7 | 73.2 | |
| VAF2026.04 | 69.8 | 94.8 | 57.6 | 71.7 | |
| VCDBase Model=LLaVA-1.5-7B2026.04 | 67.3 | — | — | 71.1 | |
| M3IDBase Model=LLaVA-1.5-7B2026.04 | 67.3 | — | — | 70.9 | |
| LLaVA-1.5-7B (Base)2026.04 | 67 | — | — | 71.1 |