Multi-Agent Path Finding on 3-agent evaluation set (3x3)
68.67Valid RateQwen3-4B + GRPO + Ours
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen3-4B + GRPO + OursBackbone=Qwen3-4B, Training=GRPO + Adaptive-weight reward2026.06 | 68.67 | 48 | |
| Kimi-K2.52026.06 | 66 | 43.33 | |
| GPT-5.42026.06 | 64.67 | 41.33 | |
| Qwen3-4B + GRPO (random)Backbone=Qwen3-4B, Training=GRPO (random)2026.06 | 54.67 | 41.33 | |
| Gemini-3.1-Pro2026.06 | 45.33 | 33.33 | |
| Qwen3-4B (base)Backbone=Qwen3-4B2026.06 | 40 | 38 | |
| Grok-4.22026.06 | 36 | 25.33 |