Visual Reasoning on MM-Star (test)
69.1AccuracyVisRef
Evaluation Results
| Method | Links | |
|---|---|---|
| VisRefModel=Qwen3-VL-8B, Configuration=Visual token selection with adaptive stopping2026.02 | 69.1 | |
| STModel=Qwen3-VL-8B, Configuration=Standard Thinking (Baseline)2026.02 | 66.5 | |
| TSRModel=Qwen3-VL-8B, Configuration=Textual Self-Reflection2026.02 | 65.9 | |
| VisRefModel=InternVL3.5-8B, Configuration=Visual token selection with adaptive stopping2026.02 | 63.1 | |
| TSRModel=InternVL3.5-8B, Configuration=Textual Self-Reflection2026.02 | 58.3 | |
| STModel=InternVL3.5-8B, Configuration=Standard Thinking (Baseline)2026.02 | 57.2 | |
| VisRefModel=SAIL-VL2-8B, Configuration=Visual token selection with adaptive stopping2026.02 | 55.3 | |
| TSRModel=SAIL-VL2-8B, Configuration=Textual Self-Reflection2026.02 | 48.9 | |
| STModel=SAIL-VL2-8B, Configuration=Standard Thinking (Baseline)2026.02 | 47.7 |