Grounding LLMs in Scientific Discovery via Embodied Actions

2026-02-24

AI Summary

Large language models (LLMs) struggle to integrate theoretical reasoning with verifiable physical simulations in scientific discovery and lack the capability for real-time perception and response to transient anomalies. To address these limitations, this work proposes EmbodiedAct, a novel framework that introduces embodied agents into scientific computing for the first time. By employing a perception-action closed-loop architecture, EmbodiedAct enables LLMs to continuously sense simulation environments and dynamically adapt their behavior in real time. Implemented in MATLAB, the framework tightly couples LLMs with scientific simulation software, significantly enhancing simulation reliability, stability, and modeling accuracy in complex engineering design and scientific modeling tasks, thereby achieving state-of-the-art performance.

Technology Category

• Intelligent Robots: Embodied AI
• Multiagent Systems: Agent-Based Simulation and Emergent Behavior
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Large Language Models (LLMs) have shown significant potential in scientific discovery but struggle to bridge the gap between theoretical reasoning and verifiable physical simulation. Existing solutions operate in a passive"execute-then-response"loop and thus lacks runtime perception, obscuring agents to transient anomalies (e.g., numerical instability or diverging oscillations). To address this limitation, we propose EmbodiedAct, a framework that transforms established scientific software into active embodied agents by grounding LLMs in embodied actions with a tight perception-execution loop. We instantiate EmbodiedAct within MATLAB and evaluate it on complex engineering design and scientific modeling tasks. Extensive experiments show that EmbodiedAct significantly outperforms existing baselines, achieving SOTA performance by ensuring satisfactory reliability and stability in long-horizon simulations and enhanced accuracy in scientific modeling.

Problem

• Large Language Models
• scientific discovery
• physical simulation
• runtime perception
• transient anomalies

Innovation

• Embodied Agents
• Perception-Execution Loop
• Scientific Discovery
• Large Language Models
• Numerical Stability