Hallucination Evaluation on CHAIR MSCOCO 2014 (val)
26.2CHAIRiDola
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DolaModel=InstructBLIP2026.02 | 26.2 | 73.2 | |
| DolaModel=LLaVA1.52026.02 | 19.3 | 61.2 | |
| DolaModel=Average2026.02 | 16.1 | 53 | |
| BeamModel=InstructBLIP2026.02 | 14.3 | 54.6 | |
| BeamModel=LLaVA1.52026.02 | 14.2 | 51.8 | |
| OPERAModel=InstructBLIP2026.02 | 14.2 | 46.4 | |
| GreedyModel=LLaVA1.52026.02 | 14.1 | 49.2 | |
| SAKEDModel=InstructBLIP2026.02 | 14 | 42.2 | |
| BeamModel=MiniGPT-42026.02 | 13.9 | 48.8 | |
| DolaModel=MiniGPT-42026.02 | 13.8 | 47.8 | |
| SAVERModel=InstructBLIP2026.02 | 13.6 | 47.8 | |
| DecoModel=InstructBLIP2026.02 | 13.5 | 46.4 | |
| GreedyModel=InstructBLIP2026.02 | 13.3 | 48.4 | |
| OPERAModel=LLaVA1.52026.02 | 12.8 | 44.6 | |
| OPERAModel=MiniGPT-42026.02 | 12.8 | 44.6 | |
| SAVERModel=LLaVA1.52026.02 | 12.8 | 40.4 | |
| DecoModel=LLaVA1.52026.02 | 12.5 | 46.4 | |
| VanillaModel=LLaVA-1.5 (13B), Decoding Strategy=Greedy2025.11 | 11.8 | 44.2 | |
| BeamModel=Average2026.02 | 11.7 | 44.6 | |
| SAKEDModel=LLaVA1.52026.02 | 11.6 | 39.6 | |
| GreedyModel=Average2026.02 | 11 | 42.8 | |
| DolaModel=Qwen2.5VL2026.02 | 11 | 41.2 | |
| RUDDERModel=LLaVA-1.5 (13B), Decoding Strategy=Greedy2025.11 | 10.8 | 39.9 | |
| DecoModel=Average2026.02 | 10.6 | 39.4 | |
| VISTAModel=LLaVA-1.5 (13B), Decoding Strategy=Greedy2025.11 | 10.6 | 40.1 | |
| SAVERModel=Average2026.02 | 10.5 | 37.4 | |
| DolaModel=InternVL32026.02 | 10.4 | 41.6 | |
| SAVERModel=MiniGPT-42026.02 | 10 | 32.8 | |
| GreedyModel=MiniGPT-42026.02 | 9.9 | 45 | |
| GreedyModel=Qwen2.5VL2026.02 | 9.8 | 40 | |
| SAKEDModel=Average2026.02 | 9.7 | 35 | |
| VanillaModel=Qwen2.5-VL (7B), Decoding Strategy=Greedy2025.11 | 9.5 | 35.2 | |
| DecoModel=MiniGPT-42026.02 | 9.4 | 32.2 | |
| DecoModel=Qwen2.5VL2026.02 | 9.4 | 38.8 | |
| BeamModel=Qwen2.5VL2026.02 | 8.4 | 36 | |
| SAVERModel=Qwen2.5VL2026.02 | 8.4 | 35.6 | |
| SAVERModel=InternVL32026.02 | 8.3 | 32.8 | |
| DecoModel=InternVL32026.02 | 8 | 33.2 | |
| GreedyModel=InternVL32026.02 | 7.8 | 31.6 | |
| SAKEDModel=MiniGPT-42026.02 | 7.7 | 30.8 | |
| VISTAModel=Qwen2.5-VL (7B), Decoding Strategy=Greedy2025.11 | 7.7 | 25.1 | |
| BeamModel=InternVL32026.02 | 7.6 | 31.6 | |
| SAKEDModel=Qwen2.5VL2026.02 | 7.6 | 32.2 | |
| SAKEDModel=InternVL32026.02 | 7.5 | 30.4 | |
| RUDDERModel=Qwen2.5-VL (7B), Decoding Strategy=Greedy2025.11 | 7 | 26.9 |