Training Large Language Models for Small-Molecule Design with Synthetic Task Scaling

• URL: http://arxiv.org/abs/2609.04735v1
• Date: Fri, 04 Sep 2026 05:05:44 GMT
• Title: Training Large Language Models for Small-Molecule Design with Synthetic Task Scaling
• Abstract summary: Large language models (LLMs) provide a useful generative prior for this problem.<n>Many chemically relevant scoring functions require hours or even days per evaluation, making them prohibitively expensive to use directly during online training.<n>We investigate whether LLMs can learn molecular design strategies from cheaper synthetic tasks that generalize to expensive molecular lead optimization settings.
• Score: 11.750407280528272
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Designing viable drug candidates requires searching a combinatorially large and rugged chemical space for molecules that satisfy multiple, often competing, objectives. Large language models (LLMs) provide a useful generative prior for this problem because of their representational capacity, reasoning ability, and flexibility when incorporating information from the external environment. While reinforcement learning from verifiable rewards (RLVR) can be used to improve the capabilities of LLMs, many chemically relevant scoring functions require hours or even days per evaluation, making them prohibitively expensive to use directly during online training. Here, we investigate whether LLMs can learn molecular design strategies from cheaper synthetic tasks that generalize to expensive molecular lead optimization settings. We find that curriculum-based training recipes that gradually incorporate more challenging synthetic design tasks enable strong performance that surpasses that of much larger frontier models on structure-based lead optimization. Our results suggest that scaling post-training using synthetic tasks is an effective strategy for adapting LLMs to high-cost experimental scenarios that are too expensive to directly train on.

Related papers

• My Chemical Harness: Evolutionary Molecular Design over Synthetic Pathways with Large Language Model Agents [2.0798270860482964]
  My Chemical Harness is a route-native evolutionary framework for goal-directed molecular design.<n>Each route is built from purchasable building blocks and reaction templates, executed by deterministic chemistry tools, and scored through task-specific molecular oracles.<n>Large language models (LLMs) are used only as strategy controllers that select high-level preferences over route length, move type, reaction families, motifs, and exploration pressure.
  arXiv  Detail & Related papers  (2026-06-08T23:52:52Z)
• Evaluating the Progression of Large Language Model Capabilities for Small-Molecule Drug Design [5.312232639944646]
  We introduce a suite of chemically-grounded tasks spanning molecular property prediction, molecular representation transformations, and molecular design.<n>We formulate these tasks as reinforcement learning environments, enabling a unified approach for evaluation and post-training.<n>A smaller model post-trained on our environments becomes competitive with state-of-the-art frontier models, despite a significantly weaker base model.
  arXiv  Detail & Related papers  (2026-04-17T17:40:26Z)
• Agentic reinforcement learning empowers next-generation chemical language models for molecular design and synthesis [51.83339196548892]
  ChemCraft is a novel framework that decouples chemical reasoning from knowledge storage.<n>ChemCraft achieves superior performance with minimal inference costs.<n>This work establishes a cost-effective and privacy-preserving paradigm for AI-aided chemistry.
  arXiv  Detail & Related papers  (2026-01-25T04:23:34Z)
• Large Language Models for Physics Instrument Design [0.0]
  We study the use of large language models (LLMs) for physics instrument design and compare their performance to reinforcement learning (RL)<n>Although RL yields stronger final designs, we find that modern LLMs consistently generate valid, resource-aware, and physically meaningful configurations.<n>In doing so, they point toward automated, closed-loop instrument design in which much of the human effort required to structure and supervise optimization can be reduced.
  arXiv  Detail & Related papers  (2026-01-12T14:30:54Z)
• Planning without Search: Refining Frontier LLMs with Offline Goal-Conditioned RL [62.984693936073974]
  Large language models (LLMs) excel in tasks like question answering and dialogue.<n>Complex tasks requiring interaction, such as negotiation and persuasion, require additional long-horizon reasoning and planning.<n>We propose a novel approach that uses goal-conditioned value functions to guide the reasoning of LLM agents.
  arXiv  Detail & Related papers  (2025-05-23T16:51:54Z)
• Cost-Optimal Grouped-Query Attention for Long-Context Modeling [45.981681856747365]
  Grouped-Query Attention (GQA) is a widely adopted strategy for reducing the computational cost of attention layers in large language models.<n>We analyze the relationship among context length, model size, GQA configuration, and model loss.<n>We propose a recipe for deriving cost-optimal GQA configurations.
  arXiv  Detail & Related papers  (2025-03-12T17:50:42Z)
• Scaling Autonomous Agents via Automatic Reward Modeling And Planning [52.39395405893965]
  Large language models (LLMs) have demonstrated remarkable capabilities across a range of tasks.<n>However, they still struggle with problems requiring multi-step decision-making and environmental feedback.<n>We propose a framework that can automatically learn a reward model from the environment without human annotations.
  arXiv  Detail & Related papers  (2025-02-17T18:49:25Z)
• MoExtend: Tuning New Experts for Modality and Task Extension [61.29100693866109]
  MoExtend is an effective framework designed to streamline the modality adaptation and extension of Mixture-of-Experts (MoE) models. MoExtend seamlessly integrates new experts into pre-trained MoE models, endowing them with novel knowledge without the need to tune pretrained models.
  arXiv  Detail & Related papers  (2024-08-07T02:28:37Z)
• Many-Shot In-Context Learning for Molecular Inverse Design [56.65345962071059]
  Large Language Models (LLMs) have demonstrated great performance in few-shot In-Context Learning (ICL) We develop a new semi-supervised learning method that overcomes the lack of experimental data available for many-shot ICL. As we show, the new method greatly improves upon existing ICL methods for molecular design while being accessible and easy to use for scientists.
  arXiv  Detail & Related papers  (2024-07-26T21:10:50Z)
• ORLM: A Customizable Framework in Training Large Models for Automated Optimization Modeling [15.67321902882617]
  We propose a viable path for training open-source LLMs capable of optimization modeling and developing solver codes.<n>This work also introduces IndustryOR, the first industrial benchmark for evaluating LLMs in solving practical OR problems.
  arXiv  Detail & Related papers  (2024-05-28T01:55:35Z)
• LMRL Gym: Benchmarks for Multi-Turn Reinforcement Learning with Language Models [56.25156596019168]
  This paper introduces the LMRL-Gym benchmark for evaluating multi-turn RL for large language models (LLMs) Our benchmark consists of 8 different language tasks, which require multiple rounds of language interaction and cover a range of tasks in open-ended dialogue and text games.
  arXiv  Detail & Related papers  (2023-11-30T03:59:31Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.