Harness optimization improves long-horizon agents by refining the prompts, tools, memory, and hooks around a frozen model. Offline methods can learn from com…
机构:Maryland
来源:arXiv 2611.05139 | AI4Papers 论文推荐平台