Benchmark performance is the dominant standard for comparing large language models (LLMs), typically summarized by a single aggregate score, such as accuracy…
机构:复旦
来源:arXiv 2610.29985 | AI4Papers 论文推荐平台