Multi-Agent Path Finding on 3-agent 10x10 (test)
32Valid RateQwen3-4B + GRPO + Ours
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-4B + GRPO + OursBackbone=Qwen3-4B, Training=GRPO + Adaptive-weight reward2026.06 | 32 | 16 | |
| Kimi-K2.52026.06 | 23.33 | 16 | |
| Qwen3-4B + GRPO (random)Backbone=Qwen3-4B, Training=GRPO (random)2026.06 | 19.33 | 13.33 | |
| Grok-4.22026.06 | 14.67 | 10 | |
| GPT-5.42026.06 | 12.67 | 7.33 | |
| Gemini-3.1-Pro2026.06 | 8 | 4.67 | |
| Qwen3-4B (base)Backbone=Qwen3-4B2026.06 | 2.67 | 2.67 |