1 paper
Zhenyu Ma, Yuyang Song, Chunyi Yang +3
LLM-based autonomous agents perform well on general reasoning tasks but still struggle to reliably use task structure, key constraints, and prior experience in complex real-world s…