ResearchBenchmarksAutonomous Agent Task Solving on AgentBoard Science WorldFollow58.9Task Completion RateReAct29.98837.4944552.506Jun 5, 2026Evaluation ResultsMethodMethodLinksTask Completion RateAverage ProgressReActLLM backbone=Claude Ha...LLM backbone=Claude Haiku 4.52026.0658.982.58AdMemLLM backbone=Claude Ha...LLM backbone=Claude Haiku 4.52026.0656.784.47AWMLLM backbone=Claude Ha...LLM backbone=Claude Haiku 4.52026.0631.163.24