Agentic large language models have been documented to lose reasoning capability when placed inside a harness, the software that mediates their observations a…
机构:MIT
来源:arXiv 2610.21218 | AI4Papers 论文推荐平台