Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Multi-Agent Pathfinding on 5-agent 5x5 (evaluation set)
Loading...
28
Valid Rate
Qwen3-4B + GRPO + LLM-as-Environment-Engineer
1.6568
8.4959
15.335
22.1741
Jun 16, 2026
Valid Rate
Optimal Rate
Updated 1mo ago
Evaluation Results
Method
Method
Links
Valid Rate
Optimal Rate
Qwen3-4B + GRPO + LLM-as-Environment-Engineer
2026.06
28
18
Grok-4.2
2026.06
26.67
16
Qwen3-4B + GRPO
training_config=random
2026.06
24
16.67
Kimi-K2.5
2026.06
23.33
17.33
GPT-5.4
2026.06
17.33
14
Gemini-3.1-Pro
2026.06
9.33
8
Qwen3-4B
variant=base
2026.06
2.67
2
Feedback
Search any
task
Search any
task