arXiv：cs.LG（机器学习，全量分类）· Xinjian Zhao, Yaoyao Xu, Xuemin Chen, Xiaozhuang Song, Tianshu Yu·· 14 小时前AI 评分31
Latent JEPA：面向化学潜在推理的抽象未来预测框架
Latent JEPA: Abstract Future Prediction for Latent Reasoning in Chemistry
AI 导读
Latent JEPA 将自回归学习与联合嵌入预测结合，可预测一个或多个未来视图，用于化学推理中的连续潜在思维训练。该框架设计了文本与分子预测目标，在 ChemCoTBench 上取得分子优化及多项编辑与反应指标的提升。表征分析显示，未来预测使潜在思维对分子结果更具信息量，并强化其与化学结构的对应关系。
正文
Abstract:Large language models offer a promising foundation for chemical reasoning, bringing together chemical knowledge and multistep problem solving. Chemical intuition can provide an initial sense of plausible outcomes before the details of a solution are fully worked out. Inspired by how such expectations complement explicit analysis, we study how continuous latent thoughts can be trained to anticipate informative aspects of future solutions without verbalizing every intermediate step. We introduce Latent JEPA, a framework that combines autoregressive learning with joint-embedding prediction of one or more future views. For chemical reasoning, we develop textual and molecular prediction objectives that connect latent thoughts to both subsequent reasoning and molecular outcomes. Experiments on ChemCoTBench show gains in molecular optimization and on several editing and reaction metrics. Representation analyses show that future prediction makes latent thoughts more informative about molecular outcomes and strengthens their correspondence with chemical structure. These findings support abstract future prediction as a learning principle for connecting continuous latent reasoning with scientific outcomes.
| Subjects: | Machine Learning (cs.LG); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.01947 [cs.LG] |
| (or arXiv:2610.01947v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.01947 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Xinjian Zhao [view email]
[v1]
Thu, 1 Oct 2026 16:11:23 UTC (6,146 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org