English

LLM-JEPA: Large Language Models Meet Joint Embedding Predictive Architectures

Computation and Language 2025-10-08 v2 Artificial Intelligence

Abstract

Large Language Model (LLM) pretraining, finetuning, and evaluation rely on input-space reconstruction and generative capabilities. Yet, it has been observed in vision that embedding-space training objectives, e.g., with Joint Embedding Predictive Architectures (JEPAs), are far superior to their input-space counterpart. That mismatch in how training is achieved between language and vision opens up a natural question: {\em can language training methods learn a few tricks from the vision ones?} The lack of JEPA-style LLM is a testimony of the challenge in designing such objectives for language. In this work, we propose a first step in that direction where we develop LLM-JEPA, a JEPA based solution for LLMs applicable both to finetuning and pretraining. Thus far, LLM-JEPA is able to outperform the standard LLM training objectives by a significant margin across models, all while being robust to overfiting. Those findings are observed across numerous datasets (NL-RX, GSM8K, Spider, RottenTomatoes) and various models from the Llama3, OpenELM, Gemma2 and Olmo families. Code: https://github.com/rbalestr-lab/llm-jepa.

Keywords

Cite

@article{arxiv.2509.14252,
  title  = {LLM-JEPA: Large Language Models Meet Joint Embedding Predictive Architectures},
  author = {Hai Huang and Yann LeCun and Randall Balestriero},
  journal= {arXiv preprint arXiv:2509.14252},
  year   = {2025}
}
R2 v1 2026-07-01T05:42:31.158Z