Toward Greater Autonomy in Materials Discovery Agents: Unifying Planning, Physics, and Scientists
🤖 AI Summary
Current language agents for crystalline materials discovery suffer from limited autonomy and excessive reliance on predefined workflows. To address this, we propose MAPPS—a highly autonomous language agent that unifies three core capabilities: automated workflow planning, physics-informed modeling (integrating foundation models for interatomic force fields), and iterative scientist-in-the-loop feedback. MAPPS enables end-to-end generation of executable code, dynamic invocation of physical models, real-time response to human feedback, and error reflection with recovery—all initiated from high-level scientific objectives. Compared to state-of-the-art methods, MAPPS achieves a fivefold improvement in stability, uniqueness, and novelty on the MP-20 dataset. Its generalizability and robustness are further validated across diverse materials discovery tasks, including crystal structure prediction, property optimization, and inverse design.
Application Category
📝 Abstract
We aim at designing language agents with greater autonomy for crystal materials discovery. While most of existing studies restrict the agents to perform specific tasks within predefined workflows, we aim to automate workflow planning given high-level goals and scientist intuition. To this end, we propose Materials Agent unifying Planning, Physics, and Scientists, known as MAPPS. MAPPS consists of a Workflow Planner, a Tool Code Generator, and a Scientific Mediator. The Workflow Planner uses large language models (LLMs) to generate structured and multi-step workflows. The Tool Code Generator synthesizes executable Python code for various tasks, including invoking a force field foundation model that encodes physics. The Scientific Mediator coordinates communications, facilitates scientist feedback, and ensures robustness through error reflection and recovery. By unifying planning, physics, and scientists, MAPPS enables flexible and reliable materials discovery with greater autonomy, achieving a five-fold improvement in stability, uniqueness, and novelty rates compared with prior generative models when evaluated on the MP-20 data. We provide extensive experiments across diverse tasks to show that MAPPS is a promising framework for autonomous materials discovery.
Problem
Research questions and friction points this paper is trying to address.
Automating workflow planning for crystal materials discovery
Unifying planning, physics, and scientist feedback in agents
Improving autonomy and reliability in materials discovery
Innovation
Methods, ideas, or system contributions that make the work stand out.
LLM-based Workflow Planner for multi-step tasks
Tool Code Generator synthesizes executable Python
Scientific Mediator ensures robustness and feedback
🔎 Similar Papers
No similar papers found.
