M3OS: A Monte Carlo Graph Search-Orchestrated Multi-Agent LLM System for Evidence-Traced Molecular Optimization

2026-09-28

AI Summary

This study addresses the limitations of large language models (LLMs) in molecular optimization, where constrained conversational contexts lead to state loss and inefficient multi-objective decision-making. To overcome these challenges, this work pioneers the integration of Monte Carlo graph search into LLM-based systems, decoupling reasoning from state management through persistent graph structures. By incorporating multi-agent collaboration, tool-driven generation, and knowledge-guided editing, the proposed framework achieves persistent optimization trajectories and separates role-specific contexts, thereby transcending conventional dialogue-based state constraints. Experimental evaluations across three benchmarks demonstrate that this approach significantly outperforms baseline methods in success rate, validating the effectiveness of persistent search states and controlled execution for multi-constrained molecular design.

Abstract

Small-molecule optimization integrates medicinal-chemistry reasoning and computational evidence through iterative, multi-objective decisions. When large language models (LLMs) reason over optimization histories stored primarily in conversational context, they must recover candidate identities, prior evaluations, and task constraints to guide subsequent decisions. We present M3OS, a multi-agent LLM system that decouples molecular-design reasoning from optimization-state management through Monte Carlo graph search. A persistent graph links evaluated candidates, parent-child transformations and evaluation evidence, while rewards and visit statistics guide LLM-assisted parent selection. Two branches combine tool-driven candidate generation with knowledge- and case-guided medicinal-chemistry editing. An execution harness controls graph updates through structured output extraction, molecular validation and task-bound evaluation. Agents receive role-specific contexts, while the graph preserves optimization trajectories beyond their active contexts. Across three molecular optimization benchmarks, M3OS achieves higher success rates than baselines, supporting the integration of persistent search state, specialized agents and controlled execution for multi-constraint optimization.

Problem

• Molecular Optimization
• Large Language Models
• Multi-Agent System
• Multi-Objective Decision
• Context Management

Innovation

• Monte Carlo Graph Search
• Multi-Agent LLM System
• Molecular Optimization
• Persistent Graph Structure
• Evidence-Traced Reasoning