Hallucination Evaluation on AMBER
24.5CHAIRmPlug-Owl
Evaluation Results
| Method | Links | ||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| mPlug-OwlLVLM=mPlug-Owl, Method=Standalone2025.09 | 24.5 | — | — | — | — | — | — | — | — | 70 | 10.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| mPlug-Owl + ORCALVLM=mPlug-Owl, Method=+ ORCA2025.09 | 18.8 | — | — | — | — | — | — | — | — | 40 | 4.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MiniGPT-4LVLM=MiniGPT-4, Method=Standalone2025.09 | 17.7 | — | — | — | — | — | — | — | — | 74 | 11.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MiniGPT-4 + ORCALVLM=MiniGPT-4, Method=+ ORCA2025.09 | 14.5 | — | — | — | — | — | — | — | — | 38 | 1.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SPARCBackbone=LLaVA-1.5-7B2026.06 | 14.5 | — | — | — | — | 77.9 | — | — | 44.6 | 23.1 | 2.4 | — | — | — | 66.7 | 65.5 | — | — | — | — | 88.5 | — | — | — | |
| AvisCModel=LLaVA-NEXT2025.05 | 14.2 | — | — | — | 77.7 | 82.1 | 88 | 76.9 | 58.1 | 65.2 | 5.7 | 84 | — | — | — | — | — | — | — | — | — | — | — | — | |
| InstructBLIPBase Model=InstructBLIP, MaxTokens=5122025.05 | 12.8 | — | — | — | — | — | — | — | 52.7 | 53.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaselineBase Model=mPLUGOwl22026.04 | 12.5 | — | — | — | — | — | — | — | 51.2 | 50.8 | 5.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDBase Model=LLaVA-NeXT, MaxTokens=5122025.05 | 12.4 | — | — | — | — | — | — | — | 61.4 | 59.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| samplingModel=LLaVA-v1.52025.05 | 12 | — | — | — | 67 | 71 | 85.2 | 60.9 | 50.3 | 51 | 4.6 | 79.5 | — | — | — | — | — | — | — | — | — | — | — | — | |
| samplingModel=LLaVA-NEXT2025.05 | 12 | — | — | — | 72.9 | 78.6 | 82.4 | 75.2 | 56.5 | 59.6 | 5.1 | 83.3 | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-1.5-7BBackbone=LLaVA-1.5-7B2026.06 | 12 | — | — | — | — | 74.7 | — | — | 50.3 | 36.4 | 4.6 | — | — | — | 65.6 | 62.4 | — | — | — | — | 83.2 | — | — | — | |
| AvisCModel=LLaVA-v1.52025.05 | 11.8 | — | — | — | 70.7 | 74.2 | 85.5 | 65.6 | 50.4 | 51.2 | 5 | 81.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDModel=LLaVA-NEXT2025.05 | 11.8 | — | — | — | 74.3 | 79.6 | 83.9 | 75.8 | 59.1 | 58.6 | 5 | 83.9 | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBase Model=mPLUGOwl22026.04 | 11.3 | — | — | — | — | — | — | — | 53.1 | 46.4 | 5.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-1.5Base Model=LLaVA-1.5, MaxTokens=5122025.05 | 11.3 | — | — | — | — | — | — | — | 50.4 | 48.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TinyLLaVA-1.5B2026.03 | 11.2 | — | — | — | — | — | — | — | 50.8 | — | 6.2 | — | 48 | — | — | — | — | — | — | — | — | — | — | — | |
| InstructBLIPBase Model=InstructBLIP, MaxTokens=AVG2025.05 | 11.2 | — | — | — | — | — | — | — | 49.6 | 44.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AvisCBase Model=LLaVA-1.5, MaxTokens=5122025.05 | 11 | — | — | — | — | — | — | — | 52.5 | 48 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBase Model=InstructBLIP, MaxTokens=5122025.05 | 10.8 | — | — | — | — | — | — | — | 53.4 | 46.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBase Model=LLaVA-NeXT, MaxTokens=5122025.05 | 10.5 | — | — | — | — | — | — | — | 63.5 | 57.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDBase Model=InstructBLIP, MaxTokens=5122025.05 | 10.4 | — | — | — | — | — | — | — | 51.7 | 47.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen-VL-ChatLVLM=Qwen-VL-Chat, Method=Standalone2025.09 | 10.4 | — | — | — | — | — | — | — | — | 34 | 3.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDModel=LLaVA-NEXT2025.05 | 10.2 | — | — | — | 75 | 80 | 84.9 | 75.7 | 59 | 52.8 | 4.6 | 84.9 | — | — | — | — | — | — | — | — | — | — | — | — | |
| AvisCBase Model=InstructBLIP, MaxTokens=5122025.05 | 10.1 | — | — | — | — | — | — | — | 51.2 | 46.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBackbone=LLaVA-1.5-7B2026.06 | 10.1 | — | — | — | — | 74.8 | — | — | 51.2 | 43.6 | 4.3 | — | — | — | 63.6 | 66.4 | — | — | — | — | 84.1 | — | — | — | |
| VCDModel=LLaVA-v1.52025.05 | 10 | — | — | — | 67.3 | 71.1 | 86.1 | 60.5 | 51.2 | 43.6 | 4.3 | 80.6 | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDBase Model=LLaVA-NeXT, MaxTokens=AVG2025.05 | 10 | — | — | — | — | — | — | — | 49.6 | 41.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GreedyModel=InstructBLIP2026.05 | 9.8 | — | — | — | 68.3 | 74.7 | 78.2 | 71.5 | 45.9 | 35.2 | 2.7 | 82.45 | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-RLHFModel=LLaVA-1.5-7B2026.06 | 9.8 | — | — | — | — | — | — | — | — | 46.4 | 5.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OPERABase Model=InstructBLIP, MaxTokens=5122025.05 | 9.7 | — | — | — | — | — | — | — | 51.2 | 40.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InstructBLIPBase Model=InstructBLIP, MaxTokens=642025.05 | 9.6 | — | — | — | — | — | — | — | 46.5 | 36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-1.5Base Model=LLaVA-1.5, MaxTokens=AVG2025.05 | 9.6 | — | — | — | — | — | — | — | 47.5 | 39.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoDModel=LLaVA-NEXT2025.05 | 9.4 | — | — | — | 78.9 | 84.3 | 83.4 | 85.2 | 61.6 | 50.8 | 3.2 | 87.5 | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-NeXTBase Model=LLaVA-NeXT, MaxTokens=5122025.05 | 9.3 | — | — | — | — | — | — | — | 60.6 | 51.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBase Model=LLaVA-NeXT, MaxTokens=AVG2025.05 | 9.3 | — | — | — | — | — | — | — | 51 | 41.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen-VL-Chat + ORCALVLM=Qwen-VL-Chat, Method=+ ORCA2025.09 | 9.3 | — | — | — | — | — | — | — | — | 24 | 0.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBase Model=InstructBLIP, MaxTokens=AVG2025.05 | 9.2 | — | — | — | — | — | — | — | 50.5 | 38.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AvisCBase Model=LLaVA-NeXT, MaxTokens=5122025.05 | 9.2 | — | — | — | — | — | — | — | 61.1 | 50.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDModel=LLaVA-v1.52025.05 | 9.1 | — | — | — | 67.3 | 70.9 | 86.5 | 60.1 | 49.4 | 42.8 | 4.4 | 80.9 | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDModel=LLaVA-v1.5-7B2025.07 | 9.1 | — | — | — | — | — | — | — | — | 39.8 | 4.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GreedyModel=LLaVA-1.52026.05 | 8.9 | — | — | — | 69.9 | 75.1 | 83.7 | 68.1 | 45 | 31.5 | 2.2 | 83.1 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen-2.5-VL2026.03 | 8.8 | — | — | — | — | — | — | — | 59.9 | — | 3.4 | — | 46.6 | — | — | — | — | — | — | — | — | — | — | — | |
| CAACBase Model=LLaVA-NeXT, MaxTokens=5122025.05 | 8.8 | — | — | — | — | — | — | — | 60.5 | 47.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaAvg. FLOPs(T)=34.322026.03 | 8.7 | — | — | — | — | — | — | — | 62.1 | 48.9 | 4.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDBase Model=InstructBLIP, MaxTokens=AVG2025.05 | 8.7 | — | — | — | — | — | — | — | 49.5 | 37.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MoDModel=LLaVA-v1.52025.05 | 8.6 | — | — | — | 72.4 | 75.3 | 91.5 | 64 | 50.7 | 38.8 | 4.7 | 83.4 | — | — | — | — | — | — | — | — | — | — | — | — | |
| HIREBase Model=TinyLLaVA-1.5B2026.03 | 8.6 | — | — | — | — | — | — | — | 48.7 | — | 2.6 | — | 35.6 | — | — | — | — | — | — | — | — | — | — | — | |
| AvisCBase Model=LLaVA-1.5, MaxTokens=AVG2025.05 | 8.6 | — | — | — | — | — | — | — | 49.5 | 36.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SHARPBackbone=LLaVA-1.5-7B2026.06 | 8.5 | — | — | — | — | — | — | — | 52.1 | 39.2 | 4.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| baselineModel=LLaVA-v1.5-7B2026.02 | 8.4 | — | — | — | — | — | — | — | — | 35.5 | 4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| baselineModel=InstructBLIP2026.03 | 8.4 | — | — | — | 68.2 | 74.6 | — | — | 46.4 | 31.1 | 2.6 | 83.1 | — | — | — | — | 2.93 | 1.04 | — | — | — | — | — | — | |
| AvisCBase Model=InstructBLIP, MaxTokens=AVG2025.05 | 8.4 | — | — | — | — | — | — | — | 49 | 37.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-RLHFData Size=122k, Feedback=Self-Reward, Model Scale=7B2026.04 | 8.4 | — | — | — | 70.2 | 76.5 | — | — | 52.3 | — | 4.6 | — | 42.5 | — | — | — | — | — | — | — | — | — | — | — | |
| baselineModel=LLaVA-v1.5-7B2025.07 | 8.4 | — | — | — | — | — | — | — | — | 35.5 | 4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RLHFBase Model=LLaVA-v1.5-7B, RL Technique=RLHF2025.07 | 8.3 | — | — | — | — | — | — | — | 52.2 | — | 4.5 | — | 41.8 | — | — | — | — | — | — | — | — | — | — | — | |
| RegularBackbone=LLaVA-1.5-7B2026.05 | 8.3 | — | — | — | 72.7 | 73.7 | — | — | — | — | — | 82.7 | — | — | — | — | — | — | — | — | — | — | — | — | |
| OPERABase Model=InstructBLIP, MaxTokens=AVG2025.05 | 8.2 | — | — | — | — | — | — | — | 48.8 | 32.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBase Model=LLaVA-1.5, MaxTokens=5122025.05 | 8.2 | — | — | — | — | — | — | — | 51.9 | 37.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDModel=LLaVA-1.5-7B2026.06 | 8.1 | — | — | — | — | — | — | — | — | 37.9 | 4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PinPointAvg. FLOPs(T)=22.182026.03 | 8 | — | — | — | — | — | — | — | 53.1 | 42.4 | 3.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| baselineModel=LLaVA-1.52026.03 | 8 | — | — | — | 67 | 71.1 | — | — | 44.5 | 31 | 2.2 | 81.58 | — | — | — | — | 9.89 | 8.07 | — | — | — | — | — | — | |
| LLaVA-1.5Base Model=LLaVA-1.5, MaxTokens=642025.05 | 8 | — | — | — | — | — | — | — | 44.5 | 31 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBase Model=LLaVA-NeXT, MaxTokens=642025.05 | 8 | — | — | — | — | — | — | — | 38.4 | 26.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Intern-VL2.5-7BModel Scale=7B2025.07 | 7.9 | — | — | — | — | — | — | — | 54.7 | — | 3.2 | — | 37.1 | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-NeXTBase Model=LLaVA-NeXT, MaxTokens=AVG2025.05 | 7.9 | — | — | — | — | — | — | — | 48.1 | 36 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaModel=LLaVA-1.5-7B2026.06 | 7.9 | — | — | — | — | — | — | — | — | 36.8 | 4.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaMLLM=LLaVA-Next2025.11 | 7.8 | — | — | — | 85.9 | 89.8 | 86.5 | 93.4 | 63.1 | 46.1 | 3.9 | 41.5 | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAFMLLM=LLaVA-Next2025.11 | 7.8 | — | — | — | 86 | 89.7 | 87.7 | 91.9 | 63.2 | 45.4 | 3.9 | 41.45 | — | — | — | — | — | — | — | — | — | — | — | — | |
| AvisCBase Model=LLaVA-NeXT, MaxTokens=AVG2025.05 | 7.8 | — | — | — | — | — | — | — | 48.6 | 35.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDModel=LLaVA-v1.5-13B2025.07 | 7.8 | — | — | — | — | — | — | — | — | 36.2 | 3.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-1.5-7BModel=LLaVA-1.5-7B, Pairs Source=Base2026.06 | 7.8 | — | — | — | — | — | — | — | — | 36.4 | — | — | — | — | 64.1 | 62.4 | — | — | — | — | 83.2 | — | — | — | |
| VCDModel=InstructBLIP2026.03 | 7.6 | — | — | — | 69.65 | 75.9 | — | — | 47.7 | 29.9 | 2.2 | 84.15 | — | — | — | — | 6.11 | 2.08 | — | — | — | — | — | — | |
| LLaVA-v1.5-7BModel Scale=7B2025.07 | 7.6 | — | — | — | — | — | — | — | 51.7 | — | 4.2 | — | 35.4 | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBase Model=InstructBLIP, MaxTokens=642025.05 | 7.6 | — | — | — | — | — | — | — | 47.5 | 29.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Muffin-13BBackbone=Muffin-13B2025.07 | 7.5 | — | — | — | — | — | — | — | 45.7 | 34.6 | 3.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VCDBase Model=LLaVA-1.5, MaxTokens=AVG2025.05 | 7.5 | — | — | — | — | — | — | — | 49.2 | 32.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDBase Model=LLaVA-NeXT, MaxTokens=642025.05 | 7.5 | — | — | — | — | — | — | — | 37.8 | 23.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AvisCModel=InstructBLIP2026.05 | 7.5 | — | — | — | 73 | 79.1 | 81.5 | 76.8 | 47 | 29.1 | 2.5 | 85.8 | — | — | — | — | — | — | — | — | — | — | — | — | |
| DolaModel=LLaVA1.52026.02 | 7.4 | — | — | — | — | 78.2 | — | — | — | — | 2.8 | — | 30.1 | — | — | — | — | — | — | — | — | — | — | — | |
| CAACBase Model=LLaVA-NeXT, MaxTokens=AVG2025.05 | 7.4 | — | — | — | — | — | — | — | 48.9 | 33.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| OPERABase Model=LLaVA-1.5, MaxTokens=5122025.05 | 7.3 | — | — | — | — | — | — | — | 47.5 | 29.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDBase Model=LLaVA-1.5, MaxTokens=5122025.05 | 7.2 | — | — | — | — | — | — | — | 57.3 | 41.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AvisCModel=LLaVA-1.52026.05 | 7.2 | — | — | — | 70.9 | 75.6 | 85.1 | 68 | 45.7 | 29.1 | 2.3 | 84.2 | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaMLLM=LLaVA-7B2025.11 | 7.1 | — | — | — | 72.1 | 74.8 | 92.2 | 63 | 50.7 | 32.5 | 3.8 | 34.35 | — | — | — | — | — | — | — | — | — | — | — | — | |
| PAIMLLM=LLaVA-Next2025.11 | 7.1 | — | — | — | 86.4 | 90.1 | 87.4 | 93 | 64.2 | 42.2 | 3.2 | 42 | — | — | — | — | — | — | — | — | — | — | — | — | |
| VGAMLLM=LLaVA-Next2025.11 | 7.1 | — | — | — | 86.6 | 90.2 | 88.4 | 92 | 62.2 | 41.4 | 3.1 | 42.05 | — | — | — | — | — | — | — | — | — | — | — | — | |
| PAIMLLM=Qwen2.5-VL2025.11 | 7.1 | — | — | — | 86 | 89.6 | 88.1 | 91.2 | 60.3 | 34.7 | 1.6 | 41.75 | — | — | — | — | — | — | — | — | — | — | — | — | |
| FLBBase Model=mPLUGOwl22026.04 | 7.1 | — | — | — | — | — | — | — | 51.6 | 33 | 2.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RLHFBackbone=Muffin-13B2025.07 | 7.1 | — | — | — | — | — | — | — | 45.2 | 37.1 | 3.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RLHFBase Model=LLaVA-v1.5-13B, RL Technique=RLHF2025.07 | 7.1 | — | — | — | — | — | — | — | 51.4 | — | 3.6 | — | 36.3 | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-1.5-7BModel Scale=7B2026.04 | 7.1 | — | — | — | 78.8 | 83 | — | — | 50.5 | — | 3.8 | — | 32.5 | — | — | — | — | — | — | — | — | — | — | — | |
| POVIDData Size=17k, Feedback=GPT-4V, Model Scale=7B2026.04 | 7.1 | — | — | — | 78.4 | 82.8 | — | — | 50.2 | — | 3.6 | — | 31.5 | — | — | — | — | — | — | — | — | — | — | — | |
| CAACBase Model=InstructBLIP, MaxTokens=5122025.05 | 7 | — | — | — | — | — | — | — | 51.9 | 30.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| baselineModel=LLaVA-v1.5-13B2026.02 | 6.9 | — | — | — | — | — | — | — | — | 31.9 | 3.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAFMLLM=LLaVA-7B2025.11 | 6.9 | — | — | — | 68.7 | 70.9 | 92.6 | 57.4 | 50.7 | 33 | 3.5 | 32.5 | — | — | — | — | — | — | — | — | — | — | — | — | |
| VAFMLLM=LLaVA-13B2025.11 | 6.9 | — | — | — | 69.5 | 71.2 | 95.6 | 56.7 | 51.2 | 32 | 3.6 | 32.65 | — | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDModel=InstructBLIP2026.03 | 6.9 | — | — | — | 69.05 | 75.25 | — | — | 47.2 | 27.5 | 2.2 | 84.2 | — | — | — | — | 5.87 | 2.08 | — | — | — | — | — | — | |
| HALVABase Model=LLaVA-v1.5-7B, RL Technique=HALVA2025.07 | 6.9 | — | — | — | — | — | — | — | 52.8 | — | 3.5 | — | 33.2 | — | — | — | — | — | — | — | — | — | — | — | |
| M3IDBase Model=InstructBLIP, MaxTokens=642025.05 | 6.9 | — | — | — | — | — | — | — | 47.2 | 27.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — |