KoLA: Carefully Benchmarking World Knowledge of Large Language Models

TL;DR

This paper introduces KoLA, a comprehensive benchmark for evaluating large language models' world knowledge, emphasizing meticulous design, diverse data sources, and innovative evaluation metrics to ensure fair and insightful assessments.

Contribution

We developed KoLA, a knowledge-oriented benchmark with a four-level ability taxonomy, diverse data sources including emerging corpora, and novel evaluation metrics for thorough LLM assessment.

Findings

28 LLMs evaluated revealing varied knowledge capabilities

KoLA's contrastive metrics provide nuanced insights into LLM knowledge creation

Benchmark and leaderboard publicly available for ongoing research

Abstract

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 \textbf{ability modeling}, we mimic human cognition to form a four-level taxonomy of knowledge-related abilities, covering $19$ tasks. (2) For \textbf{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 \textbf{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 $28$ open-source and commercial LLMs and obtain some intriguing findings. The KoLA dataset and open-participation leaderboard are publicly released at https://kola.xlore.cn and will be continuously updated to provide references for developing LLMs and knowledge-related systems.