[2603.21728] EvoIdeator: Evolving Scientific Ideas through Checklist-Grounded Reinforcement Learning

Computer Science > Artificial Intelligence arXiv:2603.21728 (cs) [Submitted on 23 Mar 2026] Title:EvoIdeator: Evolving Scientific Ideas through Checklist-Grounded Reinforcement Learning Authors:Andreas Sauter, Yuyue Zhao, Jacopo Urbani, Wenxiang Hu, Zaiqiao Meng, Lun Zhou, Xiaohui Yan, Yougang Lyu View a PDF of the paper titled EvoIdeator: Evolving Scientific Ideas through Checklist-Grounded Reinforcement Learning, by Andreas Sauter and Yuyue Zhao and Jacopo Urbani and Wenxiang Hu and Zaiqiao Meng and Lun Zhou and Xiaohui Yan and Yougang Lyu View PDF HTML (experimental) Abstract:Scientific idea generation is a cornerstone of autonomous knowledge discovery, yet the iterative evolution required to transform initial concepts into high-quality research proposals remains a formidable challenge for Large Language Models (LLMs). Existing Reinforcement Learning (RL) paradigms often rely on rubric-based scalar rewards that provide global quality scores but lack actionable granularity. Conversely, language-based refinement methods are typically confined to inference-time prompting, targeting models that are not explicitly optimized to internalize such critiques. To bridge this gap, we propose \textbf{EvoIdeator}, a framework that facilitates the evolution of scientific ideas by aligning the RL training objective with \textbf{checklist-grounded feedback}. EvoIdeator leverages a structured judge model to generate two synergistic signals: (1) \emph{lexicographic rewards} for multi-dime...