Reproduce “Re-Search: Dissecting LLM Ideation Through a Two-Player Questioning Game” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Muhan Gao, Hangxiao Zhu, Xinqi Su, Yu Zhang, Daniel Khashabi, Kuan-Hao Huang
Re-Search tests how models recover a research idea through iterative questions about a paper whose solution is hidden. Across six models and 200 ICML 2026 papers, most recovery gains arrive within the first 30 of up to 100 questions. Supplying the paper’s core idea improves scores, but this diagnostic result does not show autonomous systems can find useful ideas.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.