CORTEXA
← Browse

Xiaonan Luo

1 paper indexed

arxivcs.AIcs.CL2026-07-14

MemoHarness: Agent Harnesses That Learn from Experience

Yue Huang, Wenjie Wang, Han Bao, Yuchen Ma, Xiaonan Luo, Yi Nian, et al.

An agent harness is the external control layer that turns a base LLM into an executable agent by managing context, tools, orchestration, memory, decoding, and output handling. While harness design strongly affects agent behavior, most automatic improvement methods optimize narrow…

View free PDFSource page