IdeaScientist: Orchestrating Agents for Grounded Scientific Ideation

2026-10-02

AI Summary

This study addresses the challenge that automated scientific research struggles to generate proposals that are both promising and methodologically grounded. To overcome this limitation, the authors decompose research ideation into three sequential stages: gap discovery, innovation, and report drafting. They introduce a novel reinforcement learning-based multi-agent framework featuring division of labor, which leverages cross-domain analogical reasoning to inspire ideation. Furthermore, the work constructs the Svalbard Idea Vault corpus to facilitate cross-disciplinary insight retrieval and temporal evaluation. Empirical results demonstrate that the proposed approach, implemented on Qwen3.6-27B, outperforms the strongest baseline by 14%, exhibits significantly enhanced novelty, and surpasses several leading closed-source models in overall performance.

Abstract

Despite rapid progress in automating scientific research, generating promising and well grounded research solutions remains a central challenge. We isolate research ideation as a standalone task and build our solution on the intuition that a challenge in one field can often be addressed by a mechanism that solved an analogous challenge in another. Accordingly, we introduce IdeaScientist, which decomposes ideation into gap finding, innovation, and report writing, and trains each role with reinforcement learning. These roles identify limitations in related work, draw solution intuitions from analogous problem settings, and develop those intuitions into complete research proposals. To facilitate discovery of insights across domains, we construct the Svalbard Idea Vault, a corpus of 2.77M decomposed research ideas for retrieval, training, and temporally controlled evaluation. Our evaluation restricts access to literature available before a cutoff date and assesses how closely proposed directions align with those later explored in 15K papers authored by human researchers. On Qwen3.6-27B, IdeaScientist outperforms the strongest open-source autoresearch baseline by 14.0%, driven mainly by gains in novelty. On this 27B open backbone, IdeaScientist even outperforms Claude Code SDK with Claude-4.8-Opus and Codex SDK with GPT-5.4, by up to 5.9%.

Problem

• research ideation
• scientific discovery
• automated research
• cross-domain analogy
• idea generation

Innovation

• Multi-agent Orchestration
• Reinforcement Learning
• Scientific Ideation
• Cross-domain Analogy
• Idea Vault