AI summaryⓘ
The authors study whether large language models (LLMs) judge scientific ideas based on their actual content or just on how nicely they are written. They introduce SciStyleBench, a tool with three parts that changes the writing style of fixed scientific ideas to see if style affects scores, measures how style influences evaluations, and attempts to separate style from content during scoring. Their experiments show that LLMs are biased by writing style and often confuse style for substance, but their new method helps reduce this bias while still recognizing real scientific value. This work helps make evaluating scientific ideas fairer by focusing on the ideas themselves rather than on how they're presented.
Large language modelsStylistic biasScientific idea evaluationStyle perturbationEvaluation metricsRank robustnessStyle-content disentanglementBenchmarking
Authors
Fengxian Ji, Yuke Li, Jingpu Yang, Juanfan Wu, Fan Zhang, Zhexuan Cui, Yu Xie, Min Peng, Qianqian Xie, Xiuying Chen, Zhuohan Xie
Abstract
However, whether these judges truly evaluate the scientific substance of ideas or are influenced by superficial stylistic presentation remains an open question. To address this question, we propose SciStyleBench, a unified three-component benchmark for diagnosing and mitigating stylistic bias in LLM-based idea evaluation: (i) First, SciStyleStage, a three-stage evaluation environment that applies controlled stylistic perturbations to fixed scientific content across three settings no context, fixed-domain context, and open-domain retrieval context, covering 600 scientific ideas and 15 style variants, with 9,000 evaluation instances per setting; (ii) Second, SciStyleMetrics, a set of quantitative measures, including Style Bias Index (SBI), Substance Recognition Rate (SRR), and Adversarial Win Rate (AWR), to characterize how stylistic variation affects scoring stability, substance discrimination, and ranking robustness; (iii) Third, SciStyleExtractor, a plug-and-play evaluation module that separates presentation style from scientific content by predicting style type and deviation before style-conditioned evaluation, enabling us to assess whether style awareness reduces stylistic bias. Experiments on SciStyleBench show that direct LLM judges remain sensitive to writing style and struggle to distinguish scientific substance. In contrast, SciStyleExtractor reduces SBI from 0.566 to 0.501 while increasing SRR and AWR from 0.504 and 0.554 to 0.759 and 0.899, respectively. These results suggest that robust idea evaluation requires invariance to stylistic variation without sacrificing sensitivity to scientific substance. Overall, SciStyleBench provides a systematic framework for identifying, quantifying, and mitigating stylistic bias in scientific idea evaluation.