Sokoban Puzzle on Sokoban Base
1.89Box Placement ScoreGPT-5
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-52026.02 | 1.89 | |
| Doubao-Seed-1.82026.02 | 1.8 | |
| Doubao-Seed-1.62026.02 | 1.22 | |
| Claude-4.5-Sonnet2026.02 | 1.06 | |
| Base Model + GLAD + MC-CriticBackbone=Qwen3-4B-Instruct2026.02 | 0.94 | |
| Base Model + GLADBackbone=Qwen3-4B-Instruct2026.02 | 0.72 | |
| Qwen3-235B-A22B-Instruct2026.02 | 0.61 | |
| UITARS-1.52026.02 | 0.44 | |
| Qwen3-30B-A3B-Instruct2026.02 | 0.44 | |
| Qwen3-4B-Instruct (Base Model)2026.02 | 0.39 |