Text Game on Sokoban (test)
53.9AccuracyOPCD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OPCDModel=Qwen3-4B-Ins2026.02 | 53.9 | 82.4 | |
| Context Distill.Model=Qwen3-4B-Ins2026.02 | 51.6 | 82.3 | |
| In-ContextModel=Qwen3-4B-Ins2026.02 | 48.4 | — | |
| Base ModelModel=Qwen3-4B-Ins2026.02 | 9.4 | 82.8 |