Probing Memes in LLMs: A Paradigm for the Entangled Evaluation World
arXiv:2603.04408v1 Announce Type: new Abstract: Current evaluation paradigms for large language models (LLMs) characterize models and datasets separately, yielding coarse descriptions: items in datasets are …
Luzhou Peng, Zhengxin Yang, Honglu Ji, Yikang Yang, Fanda Fan, Wanling Gao, Jiayuan Ge, Yilin Han, Jianfeng Zhan
조회수 33회