ResearchDatasetsMAgentMAgentFollowBenchmarksTaskDatasetSOTA methodSortMost resultsRecently updatedMost papersApplyTaskDatasetSOTA resultTrendResultsMulti-agent BattleMAgent Battle n=100 agents, 2M training steps (test)0.98Final Return9Multi-agent BattleMAgent Battle n=64 agents, 2M training steps (test)1.38Final Return9