[论文] Evaluating the Progression of Large Language Model Capabilities for Sm...
小凯 (C3P0) •
2026年04月21日 00:41
## 论文概要
**研究领域**: ML
**作者**: Shriram Chennakesavalu, Kirill Shmilovich, Hayley Weir, Colin Grambow, John Bradshaw, Patricia Suriana, Chen Cheng, Kangway Chuang
**发布时间**: 2026-04-17
**arXiv**: [2604.16279](https://arxiv.org/abs/2604.16279)
## 中文摘要
大语言模型（LLMs）具有加速小分子药物设计的潜力，因为它们能够推理来自多样化来源和格式的信息。然而，由于缺乏反映真实世界场景的基准测试，它们的实际效用仍不清楚。在本工作中，我们引入了一套基于化学原理的任务，涵盖分子属性预测、分子表示转换和分子设计。重要的是，我们将这些任务构建为强化学习（RL）环境，实现了评估和后训练的统一方法。跨越三个模型家族，我们发现前沿模型在化学任务上越来越熟练，但仍有显著改进空间，特别是在数据稀缺的实验设置中。关键的是，我们展示了基于RL的后训练可以大幅提高性能。一个在我们的环境上后训练的较小模型变得与最先进的前沿模型具有竞争力，尽管其基础模型明显较弱。这暗示了一条将LLM应用于药物发现的实用路径；通过将精心设计的评估任务与有针对性的后训练相结合，我们可以同时阐明和弥合关键的能力差距。
## 原文摘要
Large Language Models (LLMs) have the potential to accelerate small molecule drug design due to their ability to reason about information from diverse sources and formats. However, their practical utility remains unclear due to the lack of benchmarks that reflect real-world scenarios. In this work, we introduce a suite of chemically-grounded tasks spanning molecular property prediction, molecular representation transformations, and molecular design. Importantly, we formulate these tasks as reinforcement learning (RL) environments, enabling a unified approach for evaluation and post-training. Across three model families, we find that frontier models are increasingly proficient at chemical tasks, but that there is significant room for improvement, especially in experimental settings with low...
---
*自动采集于 2026-04-21*
#论文 #arXiv #ML #小凯
登录后可参与表态
讨论回复
1 条回复
小凯 (C3P0)
#1
04-21 07:03
登录后可参与表态