Visual Reasoning on Sudoku (Scale-Specific Accuracy)
100Accuracy (Scale 40)EndoCoT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| EndoCoTTraining Setting=Task-Specific Training2026.03 | 100 | 95 | 64 | |
| DiffThinkerTraining Setting=Task-Specific Training2026.03 | 94 | 55 | 13 | |
| EndoCoTTraining Setting=Unified Training2026.03 | 88 | 80 | 59 | |
| DiffThinkerTraining Setting=Unified Training2026.03 | 45 | 26 | 37 | |
| Qwen3-VL-8BTraining Setting=Task-Specific Training, Optimization Method=SFT2026.03 | 17 | 2 | 0 | |
| ThinkGenTraining Setting=Zero-Shot2026.03 | 0 | 0 | 0 | |
| ChronoEditTraining Setting=Zero-Shot2026.03 | 0 | 0 | 0 | |
| Qwen3-VL-8BTraining Setting=Zero-Shot2026.03 | 0 | 0 | 0 | |
| Qwen-Image-Edit-2511Training Setting=Zero-Shot2026.03 | 0 | 0 | 0 | |
| Qwen3-VL-8BTraining Setting=Task-Specific Training, Optimization Method=GRPO2026.03 | 0 | 0 | 0 |