Aug 2026· Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2· 0 citations· 9 references
TL;DR
This paper proposes UniEdit, a Unified Graph-based Mixture-of-Experts (MoE) Molecular Editing model that offers a robust alternative to LLMs and incorporates a Mixture-of-Experts architecture that dynamically routes tasks to specialized components.
Abstract
Recently, Large Language Models (LLMs) have become the dominant paradigm for molecular editing due to their strong generalization capabilities across diverse tasks. However, treating molecules as 1D text strings (SMILES) introduces significant challenges in controllability and structural validity. In this paper, we question whether sequence generation is truly optimal for this topological task. We propose UniEdit, a Unified Graph-based Mixture-of-Experts (MoE) Molecular Editing model that offers a robust alternative to LLMs. Diverging from the generative approach, UniEdit reformulates molecular editing as a hierarchical node-level classification task. By predicting discrete edit actions (e.g., Add, Remove, Replace) directly on the graph, our model ensures topological precision by design. To handle conflicting multi-objective constraints within a single framework, we incorporate a Mixture-of-Experts architecture that dynamically routes tasks to specialized components. Extensive experiments across 28 diverse tasks demonstrate that UniEdit significantly outperforms sequence-based baselines. Furthermore, a preliminary scaling study reveals that our graph-based approach benefits consistently from increased model capacity, suggesting a scalable path toward general-purpose molecular editing. Our code is available at https://github.com/jiajunyu1999/GraphEditing.
A Molecular Perturbation framework that generates syntax-valid structural variants of training molecules under controlled Graph Edit Distance (GED) to probe the manifold regularity of molecular LLMs and suggests that it can partially expand the local trust region and offer a promising direction for stabilizing molecular LLMs against structural variation.
Jiatong Li, Weida Wang, Changmeng Zheng et al.· 0 citations
This paper introduces Caduceus, a family of MoE-enhanced foundation models built with a hierarchical pre-training paradigm to jointly integrate biological and natural language, and incorporates a multi-task instruction tuning phase, enabling robust protein parsing and natural language question answering.
Mingze Yin, Yiheng Zhu, Jialu Wu et al.· Proceedings of the 32nd ACM...· 0 citations
Automating optimization modeling from natural language with large language models (LLMs) faces two key challenges. First, training corpora lack structural diversity. Second, data generation pipelines remain static and decoupled from model learning. To address these challenges, we propose EvoOptiGraph, a novel framework where data and model co-evolve, driven by model weaknesses. EvoOptiGraph represents each mixed-integer linear program (MILP) as an attributed bipartite graph and applies validity-preserving evolutionary operators to generate structurally diverse instances. The evolved graphs are converted into solver code and natural language via deterministic compilation and verified back-translation. Training proceeds in two stages: supervised fine-tuning (SFT) on an initial dataset, followed by reinforcement learning with verifiable rewards (RLVR), where graph-derived weakness signals guide the generation of new instances targeting the model's failures. This forms a closed loop that continuously updates the training distribution. Empirical results on six public datasets show that EvoOptiGraph significantly outperforms larger generalist models, agentic methods, and specialized baselines in accuracy, executability, and generalization. These results demonstrate that targeted data-model coevolution is an effective strategy for improving LLMs on optimization modeling tasks.
Qingcan Kang, Mingyang Liu, Xiaojin Fu et al.· arXiv.org· 0 citations
Large language models (LLMs) are widely applied across chemical tasks, such as molecular property prediction, which underpins drug discovery. Molecular LLMs represent a molecule through several modalities, notably a 1D SMILES sequence or a 2D molecular graph. Both encode molecular information implicitly, so the contribution of individual substructures remains opaque. Retrieval and augmentation methods add context, but from external sources. However, the cues chemists reason over are the internal substructures that drive a property up or down. We propose MR-MoL, a multi-granular rationale-guided molecular LLM that supplies this evidence directly. A fine-tuned GNN scores each substructure through masking, and the most influential ones are serialized as a ranked, direction-tagged rationale that the LLM reads alongside the SMILES sequence and molecular graph. The rationale spans three levels of granularity: Murcko scaffolds with their side chains, BRICS fragments, and functional groups. This is, to our knowledge, the first method to expose GNN-derived attributions to an LLM as evidence for property prediction. On eight MoleculeNet tasks, MR-MoL achieves the best overall results among generalist models and narrows the gap to specialist models tuned for each task. Five diagnostics further confirm that the model reads the rationale rather than merely benefiting from its presence. Its direction, rank, and substructure each shape the prediction, and its attributions reproduce known structure-property relationships.
Junwoo Park, Minyoung Shin, C. Lee et al.· 0 citations
FragBERTa is introduced, a molecular fragment-aware transformer-based representation language model pretrained using masked language modeling on Sequential Attachment-based Fragment Embedding (SAFE) representations, suggesting that fragment-based string representations offer advantages over atom-level representations for scaffold-sensitive and interaction-driven tasks.