ThinkBench: Dynamic Out-of-Distribution Evaluation for Robust LLM Reasoning

Jun Wang (iWudao Tech) · Linyi Yang (Southern University of Science and Technology) · Weinan Zhang (Shanghai Jiao Tong University) · Yue Zhang (Suzhou University) · Kun Shao (Huawei Noah's Ark Lab) · Ying Wen (Shanghai Jiao Tong University) · Ziyu Wan (Shanghai Jiao Tong University; Bytedance Seed) · Yan Song (University College London, University of London) · Shulin Huang (Zhejiang University) · Shawn Chen (Zhejiang University) · Leyang Cui (Tencent AI Lab) · Qingcheng Zeng (Northwestern University)

contamination impactdata contamination data leakagedata reliabilitydynamic data generationevaluation frameworkexperimental conditionsnon-reasoning models out-of-distribution datasets performance evaluationprmsreasoning capability reasoning tasksrobust evaluation

Evaluating large language models (LLMs) poses significant challenges, particularly due to issues of data contamination and the leakage of correct answers. To address these challenges, we introduce ThinkBench, a novel evaluation framework designed to robustly evaluate the reasoning capability of LLMs. ThinkBench proposes a dynamic data generation method for constructing out-of-distribution (OOD) datasets and offers an OOD dataset that contains 2,912 samples drawn from reasoning tasks. ThinkBench unifies the evaluation of reasoning models and non-reasoning models. We evaluate 16 LLMs and 4 PRMs under identical experimental conditions and show that most of the LLMs' performance are far from robust and they face a certain level of data leakage. By dynamically generating OOD datasets, ThinkBench effectively provides a reliable evaluation of LLMs and reduces data contamination impact. Our data and codes are available at https://github.com/huangshulin123/ThinkBench.