Visual Reasoning on VSP-Super
100Accuracy (Scale 16)EndoCoT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EndoCoTTraining Setting=Unified Training2026.03 | 100 | 80 | |
| DiffThinkerTraining Setting=Task-Specific Training2026.03 | 99 | 80 | |
| EndoCoTTraining Setting=Task-Specific Training2026.03 | 99 | 85 | |
| DiffThinkerTraining Setting=Unified Training2026.03 | 97 | 84 | |
| Qwen3-VL-8BTraining Setting=Task-Specific Training, Optimization Method=SFT2026.03 | 61 | 8 | |
| ChronoEditTraining Setting=Zero-Shot2026.03 | 1 | 1 | |
| Qwen3-VL-8BTraining Setting=Zero-Shot2026.03 | 1 | 0 | |
| ThinkGenTraining Setting=Zero-Shot2026.03 | 0 | 0 | |
| Qwen-Image-Edit-2511Training Setting=Zero-Shot2026.03 | 0 | 0 | |
| Qwen3-VL-8BTraining Setting=Task-Specific Training, Optimization Method=GRPO2026.03 | 0 | 0 |