Sequential Decision Making on Sokoban (val)
60.4Success RateEVPO
Evaluation Results
| Method | Links | |
|---|---|---|
| EVPOBackbone=Qwen2.5-3B-Instruct2026.04 | 60.4 | |
| StarPO-SBackbone=Qwen2.5-3B-Instruct, Method Category=Critic-based2026.04 | 45.5 | |
| PPOBackbone=Qwen2.5-3B-Instruct, Method Category=Critic-based2026.04 | 31.4 | |
| DAPOBackbone=Qwen2.5-3B-Instruct, Method Category=Critic-free2026.04 | 16.2 | |
| GRPOBackbone=Qwen2.5-3B-Instruct, Method Category=Critic-free2026.04 | 15.6 | |
| Base LLMBackbone=Qwen2.5-3B-Instruct, Method Category=None2026.04 | 13.9 |