MindFlow: Mind Supernet Powered Thinking Flows for Research Idea Innovation

Research idea innovation is a fundamental engine of scientific progress, yet it remains difficult to generate and evaluate in a scalable and controllable way. This challenge lies in its inherently open-ended and multi-objective nature, where ideas should balance novelty, plausibility and feasibility. While recent LLM-based approaches have made progress through carefully designed prompts or agent pipelines, they are constrained by predefined, static ideation workflows. To address this limitation, we propose MindFlow, a framework that explicitly formulates ideation as a graph-structured Flow in Mind, which is composed of modular thinking operators and modeled by a probabilistic mind supernet. Given a research topic, a controller dynamically samples thinking flows to generate candidate ideas. This open-ended problem is optimized using a tournament-based relative ranking, enabling the controller to progressively favor higher-quality thinking flows. We further introduce an evaluation protocol that jointly assesses problem finding and problem solving, going beyond title- or abstract-only judgments. Across diverse topics, MindFlow shows its superiority as an explicit, controllable and optimizable research idea innovator.

This paper studies how agent systems can help researchers generate high-quality research ideas. A strong research idea often requires several steps, including finding an important problem, exploring possible directions, comparing alternatives, and refining a feasible solution. Existing AI systems usually follow a fixed reasoning process, which may not work well for all research topics. We propose MindFlow, a system that uses flexible thinking flows to guide research idea generation. MindFlow can adapt its reasoning process to different topics and objectives, helping it produce ideas with clearer motivations and more practical methods.