Hallucination Evaluation on ObjHal
47.3CRs AccuracyMuffin-13B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Muffin-13BBackbone=Muffin-13B2025.07 | 47.3 | — | — | 15.2 | |
| RLHFBackbone=Muffin-13B2025.07 | 45.5 | — | — | 12.7 | |
| DPOBackbone=Muffin-13B2025.07 | 43.8 | — | — | 13.9 | |
| CHiP-DPOBackbone=Muffin-13B2025.07 | 35.2 | — | — | 11.5 | |
| TARS (Replace)Backbone=Muffin-13B2025.07 | 29.3 | — | — | 8.8 | |
| TARS (Mask)Backbone=Muffin-13B2025.07 | 28.7 | — | — | 8.2 | |
| LLaVA2025.04 | — | 63 | 29.5 | — | |
| POPEN2025.04 | — | 9.2 | 4.9 | — | |
| RLHF-V2025.04 | — | 12.2 | 7.5 | — |