SciEval: A Multi-Level Large Language Model Evaluation Benchmark for Scientific Research

TL;DR

SciEval is a new multi-disciplinary benchmark designed to evaluate large language models' scientific research abilities, addressing data leakage and subjective question evaluation, revealing room for improvement especially on dynamic questions.

Contribution

It introduces SciEval, a comprehensive benchmark with dynamic and subjective questions based on Bloom's taxonomy, enhancing evaluation of LLMs' scientific research capabilities.

Findings

GPT-4 achieves state-of-the-art performance

Significant performance gap remains on dynamic questions

Benchmark addresses data leakage and subjective evaluation issues

Abstract

Recently, there has been growing interest in using Large Language Models (LLMs) for scientific research. Numerous benchmarks have been proposed to evaluate the ability of LLMs for scientific research. However, current benchmarks are mostly based on pre-collected objective questions. This design suffers from data leakage problem and lacks the evaluation of subjective Q/A ability. In this paper, we propose SciEval, a comprehensive and multi-disciplinary evaluation benchmark to address these issues. Based on Bloom's taxonomy, SciEval covers four dimensions to systematically evaluate scientific research ability. In particular, we design a "dynamic" subset based on scientific principles to prevent evaluation from potential data leakage. Both objective and subjective questions are included in SciEval. These characteristics make SciEval a more effective benchmark for scientific research ability evaluation of LLMs. Comprehensive experiments on most advanced LLMs show that, although GPT-4 achieves SOTA performance compared to other LLMs, there is still substantial room for improvement, especially for dynamic questions. The codes and data are publicly available on https://github.com/OpenDFM/SciEval.