ResearchBenchmarksLLM Agent Performance on AgentBoard BabyAI Compositional StreamFollow55.4Success Rate (1)MEMPROBE36.78441.61746.4551.283Jun 1, 2026Evaluation ResultsMethodMethodLinksSuccess Rate (1)Precision Rate (1)Success Rate (2)Precision Rate (2)MEMPROBE2026.0655.465.753.669ReMem2026.0651.868.662.575.7ReAct2026.065063.65063.6ExpRAG2026.0637.553.158.973.3