Problem
Research questions and friction points this paper is trying to address.
chemical reasoning
latent reasoning
future prediction
large language models
molecular optimization
Latent JEPA: Abstract Future Prediction for Latent Reasoning in Chemistry
🤖 AI Summary
This study addresses the issues of redundant explicit reasoning steps and weak associations between latent thought processes and molecular outcomes in chemical reasoning with large language models. To this end, we propose the Latent JEPA framework, which integrates autoregressive learning with a joint-embedding predictive architecture. By introducing dual prediction objectives over both text and molecules, the method trains a continuous latent space to anticipate future views, thereby establishing an abstract predictive mechanism that bridges latent reasoning with scientific outcomes. Experimental results demonstrate that the proposed framework significantly improves molecular optimization performance on ChemCoTBench, effectively enhancing the capacity of latent representations to capture molecular structural information.
📝 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.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Latent JEPA
Joint-Embedding Prediction
Chemical Reasoning
Latent Thoughts
Abstract Future Prediction
🔎 Similar Papers
No similar papers found.
