Problem Solving and Unsolvability Detection on Maze Easy
100Accuracy (Solvable)Deepseek-V3.2-R
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Deepseek-V3.2-RModel Scale=V3.2-R2025.12 | 100 | 98.9 | 99.5 | |
| Gemini-3Model Scale=32025.12 | 99 | 94.6 | 96.8 | |
| Qwen3-4B + UnsolvableRLModel Scale=4B, Training Protocol=UnsolvableRL2025.12 | 96.5 | 98.9 | 97.7 | |
| GPT-5.1-LowModel Scale=5.1-Low2025.12 | 95 | 100 | 97.5 | |
| Qwen3-4B InstructModel Scale=4B, Training Protocol=Instruct2025.12 | 32.7 | 55.6 | 44.1 | |
| Qwen3-1.7B InstructModel Scale=1.7B, Training Protocol=Instruct2025.12 | 0 | 85.6 | 42.8 | |
| Qwen3-1.7B + UnsolvableRLModel Scale=1.7B, Training Protocol=UnsolvableRL2025.12 | 0 | 95.2 | 47.6 |