Large language models (LLMs) excel across diverse tasks, benefiting from knowledge priors, in-context reasoning, and test-time computation. Despite these str…
机构:上交
来源:arXiv 2610.12927 | AI4Papers 论文推荐平台