Share your thoughts, 1 month free Claude Pro on usSee more

Multi-Agent Path Finding

Benchmarks

Dataset Name	SOTA Method	Metric
Maze Map	Single Policy Oracle (SPO)	AUC Ratio (SPO/Baseline)1.28	98	4mo ago
Random Map	Single Policy Oracle (SPO)	AUC Ratio (SPO / Best Baseline)1.47	98	4mo ago
Room Map	Single Policy Oracle (SPO)	AUC Ratio1.21	98	4mo ago
Large Maze 33x33 world size, 33.0% static obstacle rate	DiffLNS	Success Rate100	20	2mo ago
Medium Warehouse 25x25 world size, 34.6% static obstacle rate	DiffLNS	Success Rate100	20	2mo ago
Medium Room 23x23 world size, 33.5% static obstacle rate	DiffLNS	Success Rate100	20	2mo ago
Medium Maze 25x25 world size, 32.8% static obstacle rate	DiffLNS	Success Rate100	20	2mo ago
Small Random 10x10 world size, 17.5% static obstacle rate	DiffLNS	Success Rate96	20	2mo ago
400x400x25 grid environment (4 temporal NFZs) (20 scenarios)	DTAPP-IICR	Success Rate1	20	5mo ago
400x400x25 grid environment 2 temporal NFZs (20 scenarios)	DTAPP-IICR	Success Rate1	20	5mo ago
400x400x25 grid environment 0 temporal NFZs (20 scenarios)	DTAPP-IICR	Success Rate100	20	5mo ago
Random Map 240x240, 0.3 density	PC2P	Success Rate (SR)98	15	5mo ago
Random Map 120x120, 0.3 density	PC2P	Success Rate100	15	5mo ago
random-32-32-20 map (test)	CMAS	Success Rate (SR)100	12	4mo ago
3-agent Sum (evaluation set)	Qwen3-4B + GRPO + Ours	Valid Rate51.67	7	1mo ago
3-agent 10x10 (test)	Qwen3-4B + GRPO + Ours	Valid Rate32	7	1mo ago
3-agent 9x9 (test)	Qwen3-4B + GRPO + Ours	Valid Rate44	7	1mo ago
3-agent 8x8 (test)	Qwen3-4B + GRPO + Ours	Valid Rate42.67	7	1mo ago
3-agent 7x7 (evaluation)	Qwen3-4B + GRPO + Ours	Valid Rate46	7	1mo ago
3-agent 6x6 (evaluation set)	Qwen3-4B + GRPO + Ours	Valid Rate52.67	7	1mo ago
3-agent 5x5 (evaluation set)	Qwen3-4B + GRPO + Ours	Valid Rate62.67	7	1mo ago
3-agent 4x4 (test)	Qwen3-4B + GRPO + Ours	Valid Rate64.67	7	1mo ago
3-agent evaluation set (3x3)	Qwen3-4B + GRPO + Ours	Valid Rate68.67	7	1mo ago
MAPF-FrozenLake 4-agent Sum	Qwen3-4B + GRPO + LLM-as-Environment-Engineer	Valid Rate33.14	7	1mo ago
MAPF-FrozenLake 4-agent 10x10	Qwen3-4B + GRPO + LLM-as-Environment-Engineer	Valid Rate18.67	7	1mo ago

Showing 25 of 64 rows