From Tokens to Blocks: A Block-Diffusion Perspective on Molecular Generation

2026-01-29

AI Summary

Existing molecular language models struggle to effectively capture molecular graph structures and lack the capability for targeted generation. To address these limitations, this work proposes SoftMol, a novel framework that introduces rule-free soft fragment SMILES representations and a block diffusion language model (SoftBD), which uniquely integrates bidirectional diffusion with autoregressive generation mechanisms. Furthermore, SoftMol incorporates a gated Monte Carlo tree search strategy to enable goal-directed molecular design. Evaluated on the ZINC-Curated dataset, SoftMol achieves 100% chemical validity, improves binding affinity by 9.7%, enhances molecular diversity by 2–3 times, and accelerates inference efficiency by 6.6-fold compared to existing approaches.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Diffusion Models for Vision
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Drug discovery can be viewed as a combinatorial search over an immense chemical space, motivating the development of deep generative models for de novo molecular design. Among these, GPT-based molecular language models (MLM) have shown strong molecular design performance by learning chemical syntax and semantics from large-scale data. However, existing MLMs face two fundamental limitations: they inadequately capture the graph-structured nature of molecules when formulated as next-token prediction problems, and they typically lack explicit mechanisms for target-aware generation. Here, we propose SoftMol, a unified framework that co-designs molecular representation, model architecture, and search strategy for target-aware molecular generation. SoftMol introduces soft fragments, a rule-free block representation of SMILES that enables diffusion-native modeling, and develops SoftBD, the first block-diffusion molecular language model that combines local bidirectional diffusion with autoregressive generation under molecular structural constraints. To favor generated molecules with high drug-likeness and synthetic accessibility, SoftBD is trained on a carefully curated dataset named ZINC-Curated. SoftMol further integrates a gated Monte Carlo tree search to assemble fragments in a target-aware manner. Experimental results show that, compared with current state-of-the-art models, SoftMol achieves 100% chemical validity, improves binding affinity by 9.7%, yields a 2-3x increase in molecular diversity, and delivers a 6.6x speedup in inference efficiency. Code is available at https://github.com/szu-aicourse/softmol

Problem

• molecular generation
• graph-structured molecules
• target-aware generation
• molecular language models
• chemical space

Innovation

• block diffusion
• soft fragments
• target-aware generation
• molecular language model
• Monte Carlo tree search