Hallucination on MMVP
72.1AccuracyQwen2.5-VL
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-VLSize=72B2026.02 | 72.1 | |
| Visual Para-ThinkerSize=7B2026.02 | 71.3 | |
| Gemini2.5-ProSize=-2026.02 | 69.8 | |
| Majority voting@4Size=7B2026.02 | 68.9 | |
| SequentialSize=7B2026.02 | 68.7 | |
| Qwen2.5-VLSize=7B2026.02 | 68.3 | |
| GPT5-miniSize=-2026.02 | 65.3 | |
| Claude-4-SonnetSize=-2026.02 | 63.9 | |
| GPT-4oSize=-2026.02 | 63.8 | |
| Visual Para-ThinkerSize=3B2026.02 | 63.6 | |
| Majority voting@4Size=3B2026.02 | 60.1 | |
| SequentialSize=3B2026.02 | 60 | |
| Qwen2.5-VLSize=3B2026.02 | 59.2 |