Sequential Planning on Maze
100Score (L=8)DiffThinker
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DiffThinkerSetting=Flow Matching2025.12 | 100 | 97 | 56 | |
| DiffThinker++Setting=Flow Matching2025.12 | 100 | 100 | 65 | |
| Qwen3-VL-32BSetting=SFT2025.12 | 91 | 57 | 3 | |
| Qwen3-VL-8BSetting=SFT2025.12 | 53 | 37 | 0 | |
| GPT-5Setting=N/A2025.12 | 2 | 0 | 0 | |
| Gemini-3-FlashSetting=N/A2025.12 | 0 | 0 | 0 | |
| Qwen3-VL-8BSetting=N/A2025.12 | 0 | 0 | 0 | |
| Qwen3-VL-8BSetting=GRPO2025.12 | 0 | 0 | 0 | |
| Qwen3-VL-32BSetting=N/A2025.12 | 0 | 0 | 0 | |
| Qwen3-VL-32BSetting=GRPO2025.12 | 0 | 0 | 0 | |
| Qwen-Image-Edit-2509Setting=N/A2025.12 | 0 | 0 | 0 | |
| Qwen-Image-Edit-2511Setting=N/A2025.12 | 0 | 0 | 0 |