Skip to yearly menu bar Skip to main content


KoLA: Carefully Benchmarking World Knowledge of Large Language Models

Jifan Yu · Xiaozhi Wang · Shangqing Tu · Shulin Cao · Daniel Zhang-Li · Xin Lv · Hao Peng · Zijun Yao · Xiaohan Zhang · Hanming Li · Chunyang Li · Zheyuan Zhang · Yushi Bai · Yantao Liu · Amy Xin · Kaifeng Yun · Linlu Gong · Nianyi Lin · Jianhui Chen · Zhili Wu · Yunjia Qi · Weikai Li · Yong Guan · Kaisheng Zeng · Ji Qi · Hailong Jin · Jinxin Liu · Yu Gu · Yuan Yao · Ning Ding · Lei Hou · Zhiyuan Liu · Xu Bin · Jie Tang · Juanzi Li

Halle B #198
[ ]
Tue 7 May 7:30 a.m. PDT — 9:30 a.m. PDT


The unprecedented performance of large language models (LLMs) necessitates improvements in evaluations. Rather than merely exploring the breadth of LLM abilities, we believe meticulous and thoughtful designs are essential to thorough, unbiased, and applicable evaluations. Given the importance of world knowledge to LLMs, we construct a Knowledge-oriented LLM Assessment benchmark (KoLA), in which we carefully design three crucial factors: (1) For ability modeling, we mimic human cognition to form a four-level taxonomy of knowledge-related abilities, covering 19 tasks. (2) For data, to ensure fair comparisons, we use both Wikipedia, a corpus prevalently pre-trained by LLMs, along with continuously collected emerging corpora, aiming to evaluate the capacity to handle unseen data and evolving knowledge. (3) For evaluation criteria, we adopt a contrastive system, including overall standard scores for better numerical comparability across tasks and models, and a unique self-contrast metric for automatically evaluating knowledge-creating ability. We evaluate 21 open-source and commercial LLMs and obtain some intriguing findings. The KoLA dataset will be updated every three months to provide timely references for developing LLMs and knowledge-related systems.

Chat is not available.