Object Hallucination Detection on POPE popular
86.59F1 ScoreOPERA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OPERALVLM backbone=LLaVA-1.52026.06 | 86.59 | — | |
| RSPModel=LLaVA-1.5, Routing Layer=L232026.05 | 86.5 | 5 | |
| BaselineModel=LLaVA-1.52026.05 | 86.4 | — | |
| FoxLVLM backbone=LLaVA-1.52026.06 | 86.3 | — | |
| Always-onModel=LLaVA-1.52026.05 | 86.2 | — | |
| CausalMMLVLM backbone=LLaVA-1.52026.06 | 86.17 | — | |
| SIDLVLM backbone=LLaVA-1.52026.06 | 85.21 | — | |
| RSPModel=InstructBLIP, Prompting Strategy=Selective Prompting (RSP), Routing Mechanism=inverse top-1 confidence, Prompt Text=Be careful.2026.05 | 85.2 | 14 | |
| Always-onModel=InstructBLIP, Prompting Strategy=Always-on, Prompt Text=Be careful.2026.05 | 84.2 | — | |
| ICDLVLM backbone=LLaVA-1.52026.06 | 84.18 | — | |
| BaselineModel=InstructBLIP2026.05 | 83.9 | — | |
| FoxLVLM backbone=InstructBLIP2026.06 | 83.77 | — | |
| SIDLVLM backbone=InstructBLIP2026.06 | 83.51 | — | |
| SamplingLVLM backbone=LLaVA-1.52026.06 | 83.4 | — | |
| CausalMMLVLM backbone=InstructBLIP2026.06 | 83.27 | — | |
| OPERALVLM backbone=Shikra2026.06 | 83.21 | — | |
| OPERALVLM backbone=InstructBLIP2026.06 | 83.19 | — | |
| CausalMMLVLM backbone=Shikra2026.06 | 82.89 | — | |
| FoxLVLM backbone=Shikra2026.06 | 82.27 | — | |
| VCDLVLM backbone=LLaVA-1.52026.06 | 82.05 | — | |
| ICDLVLM backbone=InstructBLIP2026.06 | 81.74 | — | |
| SamplingLVLM backbone=InstructBLIP2026.06 | 81.55 | — | |
| VCDLVLM backbone=InstructBLIP2026.06 | 81.12 | — | |
| VCDLVLM backbone=Shikra2026.06 | 81.12 | — | |
| SamplingLVLM backbone=Shikra2026.06 | 81.06 | — | |
| ICDLVLM backbone=Shikra2026.06 | 80.2 | — | |
| SIDLVLM backbone=Shikra2026.06 | 79.97 | — |