Spatial Decision-Making on Maze size 8
92.71Success RateGemini-3-Flash
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-3-FlashSetting=Naive2026.06 | 92.71 | |
| SpecFlowSetting=Flow(image) + AR(Text)2026.06 | 92.66 | |
| SpecFlowSetting=Flow(image) + AR(Text)2026.06 | 92.04 | |
| Gemini-3-FlashSetting=Naive2026.06 | 91.12 | |
| DiffThinkerSetting=Flow(image) Only2026.06 | 90.82 | |
| DiffThinkerSetting=Flow(image) Only2026.06 | 89.95 | |
| GPT-5.1Setting=Naive2026.06 | 89.82 | |
| MVoTSetting=Autoregressive2026.06 | 89.14 | |
| GPT-5.1Setting=Naive2026.06 | 88.94 | |
| MVoTSetting=Autoregressive2026.06 | 88.61 | |
| Qwen3-VL-32BSetting=SFT2026.06 | 87.43 | |
| Qwen3-VL-32BSetting=GRPO2026.06 | 84.18 | |
| Qwen3-VL-8BSetting=SFT2026.06 | 80.34 | |
| Qwen3-VL-8BSetting=GRPO2026.06 | 76.85 | |
| Qwen3-VL-32BSetting=Naive2026.06 | 74.05 | |
| Qwen3-VL-32BSetting=SFT2026.06 | 70.84 | |
| Qwen3-VL-32BSetting=GRPO2026.06 | 67.31 | |
| Qwen3-VL-8BSetting=Naive2026.06 | 63.74 | |
| Qwen3-VL-8BSetting=SFT2026.06 | 61.47 | |
| Qwen3-VL-8BSetting=GRPO2026.06 | 57.88 | |
| Qwen3-VL-32BSetting=Naive2026.06 | 54.36 | |
| Qwen3-VL-8BSetting=Naive2026.06 | 42.75 |