Enhancing Spatial Reasoning in Large Language Models for Metal-Organic Frameworks Structure Prediction

TL;DR

MOF-LLM addresses the challenging problem of predicting 3D MOF structures by employing a block-level autoregressive framework powered by a three-stage training pipeline: spatial-aware continual pre-training, structural supervised fine-tuning, and matching-driven reinforcement learning. This approach injects explicit geometric priors, enables precise roto-translations through Euler-angle rotations, and refines structure stability via SAPO. On a large MOF dataset, MOF-LLM achieves a strong unseen-MOF match rate while maintaining rapid, single-pass inference, outperforming both denoising-based CSP models and prior LLM-based methods. The work demonstrates the practical potential of LLMs for rapid, scalable MOF discovery and design, while outlining avenues for handling block flexibility and improved 3D geometry tokenization in future efforts.

Abstract

Metal-organic frameworks (MOFs) are porous crystalline materials with broad applications such as carbon capture and drug delivery, yet accurately predicting their 3D structures remains a significant challenge. While Large Language Models (LLMs) have shown promise in generating crystals, their application to MOFs is hindered by MOFs' high atomic complexity. Inspired by the success of block-wise paradigms in deep generative models, we pioneer the use of LLMs in this domain by introducing MOF-LLM, the first LLM framework specifically adapted for block-level MOF structure prediction. To effectively harness LLMs for this modular assembly task, our training paradigm integrates spatial-aware continual pre-training (CPT), structural supervised fine-tuning (SFT), and matching-driven reinforcement learning (RL). By incorporating explicit spatial priors and optimizing structural stability via Soft Adaptive Policy Optimization (SAPO), our approach substantially enhances the spatial reasoning capability of a Qwen-3 8B model for accurate MOF structure prediction. Comprehensive experiments demonstrate that MOF-LLM outperforms state-of-the-art denoising-based and LLM-based methods while exhibiting superior sampling efficiency.

Figures (5)

• Figure 1: Illustration of MOF structures. An MOF structure and its unit cell are shown on top. The unit cell can be decomposed into building blocks, including metal nodes and organic linkers (bottom). Atom color: Zn (purple), O (red), C (gray), H (white).
• Figure 2: Overview of our MOF-LLM framework. It enhances the spatial reasoning ability of LLMs for MOF structure prediction via a three-stage training: spatial-aware CPT to inject spatial and geometric priors, structural SFT enables block assembly, and matching-driven RL refines structure stability using Soft Adaptive Policy Optimization (SAPO).
• Figure 3: Ablation study of key design components. CPT (w/o spatial info) denotes CPT using prompts without topology codes, molecular weight, PCA spans, and rotated axes.
• Figure 4: Base CPT prompt template.
• Figure 5: SFT prompt template.