Discriminative Object Hallucination on POPE MSCOCO Adversarial
87.33AccuracyYARD
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| YARDBase Model=InstructBLIP2026.05 | 87.33 | 86.87 | — | — | |
| IFCDVLM=LLaVA-1.52026.04 | 85.17 | 84.84 | — | — | |
| CMPCDecoding Strategy=Greedy, Backbone=LLaVA-OV2025.01 | 84.93 | 84.67 | — | — | |
| VCDDecoding Strategy=Greedy, Backbone=LLaVA-OV2025.01 | 84.2 | 83.86 | — | — | |
| BaselineDecoding Strategy=Greedy, Backbone=LLaVA-OV2025.01 | 84.17 | 84.02 | — | — | |
| IECD2VLM=LLaVA-1.52026.04 | 84.1 | 83.03 | — | — | |
| CMPCDecoding Strategy=Sampling, Backbone=LLaVA-OV2025.01 | 83.69 | 83.4 | — | — | |
| IECD2VLM=InstructBLIP2026.04 | 83.23 | 82.94 | — | — | |
| CMPCBase Model=LLaVA 1.52025.01 | 83.23 | 82.73 | 85.28 | 80.33 | |
| VaLiDVLM=LLaVA-1.52026.04 | 83.2 | 83.29 | — | — | |
| CATCHVLM=LLaVA-1.52026.04 | 83.17 | 83.18 | — | — | |
| VCDDecoding Strategy=Sampling, Backbone=LLaVA-OV2025.01 | 82.89 | 82.38 | — | — | |
| OctopusVLM=InstructBLIP2026.04 | 82.83 | 81.43 | — | — | |
| YARDBase Model=LLaVA-1.5-7B2026.05 | 82.73 | 82.9 | — | — | |
| CMPCBase Model=InstructBLIP2025.01 | 82.57 | 81.67 | 86.1 | 77.67 | |
| AVISCBase Model=LLaVA-1.5-7B2026.05 | 82.57 | 81.48 | — | — | |
| VCDBase Model=LLaVA-1.5-7B2026.05 | 82.33 | 81.53 | — | — | |
| IFCDVLM=InstructBLIP2026.04 | 82.23 | 80.44 | — | — | |
| OctopusVLM=LLaVA-1.52026.04 | 82.22 | 81.44 | — | — | |
| EVASBase Model=LLaVA-1.5-7B2026.05 | 82.03 | 80.74 | — | — | |
| Object-aligned VCDVLM=LLaVA-1.52026.04 | 81.9 | 82 | — | — | |
| LLaVA-1.5-7BBase Model=LLaVA-1.5-7B2026.05 | 81.51 | 80.54 | — | — | |
| VaLiDVLM=InstructBLIP2026.04 | 81.33 | 79.59 | — | — | |
| ICDBase Model=LLaVA-1.5-7B2026.05 | 81.2 | 80.45 | — | — | |
| ICDBase Model=LLaVA 1.52025.01 | 81.17 | 79.52 | 87.13 | 73.13 | |
| VCDBase Model=InstructBLIP2026.05 | 81.03 | 80.57 | — | — | |
| BaselineDecoding Strategy=Sampling, Backbone=LLaVA-OV2025.01 | 80.99 | 80.9 | — | — | |
| AVISCBase Model=InstructBLIP2026.05 | 80.9 | 80.14 | — | — | |
| VCDVLM=LLaVA-1.52026.04 | 80.88 | 81.33 | — | — | |
| VCDBase Model=LLaVA 1.52025.01 | 80.88 | 81.33 | 79.45 | 83.29 | |
| ICDBase Model=InstructBLIP2025.01 | 80.87 | 80.84 | 80.95 | 80.73 | |
| ICDVLM=LLaVA-1.52026.04 | 80.23 | 78.52 | — | — | |
| M3IDBase Model=LLaVA-1.5-7B2026.05 | 79.93 | 78.5 | — | — | |
| CATCHVLM=InstructBLIP2026.04 | 79.9 | 81.37 | — | — | |
| ICDVLM=InstructBLIP2026.04 | 79.56 | 79.52 | — | — | |
| VCDVLM=InstructBLIP2026.04 | 79.56 | 79.52 | — | — | |
| VCDBase Model=InstructBLIP2025.01 | 79.56 | 79.52 | 79.67 | 79.39 | |
| InstructBLIPBase Model=InstructBLIP2026.05 | 78.98 | 79.43 | — | — | |
| RegularVLM=LLaVA-1.52026.04 | 78.96 | 77.57 | — | — | |
| BaselineBase Model=LLaVA 1.52025.01 | 78.96 | 77.57 | 83.06 | 72.75 | |
| MFCDVLM=LLaVA-1.52026.04 | 77.03 | 79.38 | — | — | |
| RegularVLM=InstructBLIP2026.04 | 75.84 | 76.59 | — | — | |
| BaselineBase Model=InstructBLIP2025.01 | 75.84 | 76.59 | 74.3 | 79.03 |