PiFlow：基於多智能體協作的原理感知科學發現

PiFlow: Principle-aware Scientific Discovery with Multi-Agent Collaboration

May 21, 2025

作者: Yingming Pu, Tao Lin, Hongyu Chen

cs.AI

摘要

基於大型語言模型（LLM）的多智能體系統（MAS）展現出科學發現的顯著潛力。然而，現有方法通常使用缺乏合理性約束的預定義工作流程來自動化科學發現，這往往導致無目的的假設生成，並未能始終如一地將假設與證據聯繫起來，從而阻礙了系統性的不確定性降低。克服這些限制根本上需要系統性的不確定性降低。我們引入了PiFlow，這是一個信息理論框架，將自動化科學發現視為一個由原則（例如科學定律）指導的結構化不確定性降低問題。在三個不同的科學領域——發現具有目標特性的奈米材料結構、生物分子和超導體候選材料——的評估中，我們的方法顯著提高了發現效率，這反映在特性值與探索步驟的曲線下面積（AUC）增加了73.55%，並且與基礎智能體系統相比，解決方案質量提升了94.06%。總體而言，PiFlow作為一種即插即用的方法，建立了高效自動化科學發現的新範式轉變，為更穩健和加速的AI驅動研究鋪平了道路。代碼已在我們的GitHub上公開：https://github.com/amair-lab/PiFlow。

English

Large Language Model (LLM)-based multi-agent systems (MAS) demonstrate remarkable potential for scientific discovery. Existing approaches, however, often automate scientific discovery using predefined workflows that lack rationality constraints. This often leads to aimless hypothesizing and a failure to consistently link hypotheses with evidence, thereby hindering systematic uncertainty reduction. Overcoming these limitations fundamentally requires systematic uncertainty reduction. We introduce PiFlow, an information-theoretical framework, treating automated scientific discovery as a structured uncertainty reduction problem guided by principles (e.g., scientific laws). In evaluations across three distinct scientific domains -- discovering nanomaterial structures, bio-molecules, and superconductor candidates with targeted properties -- our method significantly improves discovery efficiency, reflected by a 73.55\% increase in the Area Under the Curve (AUC) of property values versus exploration steps, and enhances solution quality by 94.06\% compared to a vanilla agent system. Overall, PiFlow serves as a Plug-and-Play method, establishing a novel paradigm shift in highly efficient automated scientific discovery, paving the way for more robust and accelerated AI-driven research. Code is publicly available at our https://github.com/amair-lab/PiFlow{GitHub}.