Multimodal Judgment on MM-Vet
37.4Overall ScorePerception-Judge-Flex
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Perception-Judge-FlexSize=7B2026.06 | 37.4 | 48.2 | 65.2 | 43.5 | |
| GPT-4VLoRA Training=true2026.06 | 36.7 | 51.7 | 65.4 | 41.1 | |
| Qwen3-VL-ThinkingSize=8B2026.06 | 35.2 | 51.2 | 71.2 | 47 | |
| GPT-4oLoRA Training=true2026.06 | 34.2 | 51.7 | 66 | — | |
| Perception-Judge-Qwen3Size=4B2026.06 | 33.2 | 50.4 | 71.7 | 43.4 | |
| Flex-Judge-VLSize=32B, LoRA Training=true2026.06 | 33 | 52.9 | 67.7 | 43.9 | |
| LLaVA-1.5Size=13B2026.06 | 32.2 | 25.4 | 34.4 | 65.2 | |
| Flex-Judge-VLSize=7B2026.06 | 31.1 | 50.7 | 63 | 47.8 | |
| Perception-Judge-FlexSize=32B, LoRA Training=true2026.06 | 31 | 49.8 | 68.6 | 43.9 | |
| Qwen3-VL-ThinkingSize=4B2026.06 | 30.8 | 49.3 | 65.6 | 47 | |
| LLaVA-CriticSize=7B, LoRA Training=true2026.06 | 27.2 | 46.4 | 63.5 | — | |
| Perception-Judge-Qwen3Size=8B2026.06 | 27 | 51.4 | 74.7 | 43.5 | |
| LLaVA-1.6Size=34B2026.06 | 26.5 | — | — | — | |
| Gemini-1.0-Pro-Vision2026.06 | 20.7 | 31.2 | 40 | 52.9 | |
| Prometheus-VSize=13B, LoRA Training=true2026.06 | 15.7 | — | — | — | |
| Qwen2.5-VL-InstructSize=7B2026.06 | 10.3 | 42 | 38.8 | 57.2 |