MSP-LLM: A Unified Large Language Model Framework for Complete Material Synthesis Planning

TL;DR

MSP-LLM tackles the complete Material Synthesis Planning problem by decomposing it into precursor prediction (PP) and synthesis operation prediction (SOP), and by introducing a discrete material class variable $G$ to create a disciplined decision chain. It uses independently fine-tuned LLMs for PP and SOP, augmented with hierarchical precursor types and a precursor constraint factorization (PCF) to explicitly preserve precursor information in decoding, enabling coherent end-to-end MSP from target compositions. Experiments on real inorganic synthesis data show that MSP-LLM outperforms baselines on PP, SOP, and the full MSP task, with the complete MSP Top-10 accuracy reaching 18.56% (Split 1) and 23.23% (Split 2). An information-theoretic analysis clarifies why PCF reduces irreducible uncertainty and improves precursor-aware operation generation, supporting the framework’s robustness and practical potential for accelerating materials discovery.

Abstract

Material synthesis planning (MSP) remains a fundamental and underexplored bottleneck in AI-driven materials discovery, as it requires not only identifying suitable precursor materials but also designing coherent sequences of synthesis operations to realize a target material. Although several AI-based approaches have been proposed to address isolated subtasks of MSP, a unified methodology for solving the entire MSP task has yet to be established. We propose MSP-LLM, a unified LLM-based framework that formulates MSP as a structured process composed of two constituent subproblems: precursor prediction (PP) and synthesis operation prediction (SOP). Our approach introduces a discrete material class as an intermediate decision variable that organizes both tasks into a chemically consistent decision chain. For OP, we further incorporate hierarchical precursor types as synthesis-relevant inductive biases and employ an explicit conditioning strategy that preserves precursor-related information in the autoregressive decoding state. Extensive experiments show that MSP-LLM consistently outperforms existing methods on both PP and SOP, as well as on the complete MSP task, demonstrating an effective and scalable framework for MSP that can accelerate real-world materials discovery.

Figures (3)

• Figure 1: Overall framework of MSP-LLM. Given a target material, MSP-LLM first predicts the precursor set (PP), and then, conditioned on the predicted precursors, generates the synthesis operation sequence (SOP), resulting in a complete MSP. The input–output texts above the PP and SOP LLMs illustrate example prompts used during LLM fine-tuning for each task.
• Figure 2: SOP performance under target-only, implicit, and explicit conditioning across two fine-tuned backbone LLMs.
• Figure 3: Effect of the material group decision chain on performance(NED) across 11 material classes.