Joint-Embedding Predictive Architecture

Also known as JEPA

Joint-Embedding Predictive Architectures learn useful representations by predicting a target encoder’s latent features from visible or past context, rather than reconstructing every input detail. Common approaches use masked context-target prediction, online or EMA target encoders, and regularizers such as SIGReg to prevent collapse; action-conditioned variants model latent dynamics for planning. These methods are being applied to images, video, robotics, medical and scientific data, and time series, with evaluations ranging from linear probes to control and forecasting benchmarks. A central open tension is that predictable, non-collapsed features are not necessarily the features needed for a task: work tests how to preserve control-relevant or diagnostic information and how well prediction metrics reflect downstream performance.

Latest papers

249 papers60 papers in the last four weeks, up 88% on the four weeks before. 0.3% of all new papers.

Proving and Improving the Platonic Representation Hypothesis with JEPAs
Oct 29, 2026Randall Balestriero, Yann LeCunPlatonic Representation Hypothesis Cross-Modal Alignment
Brown University · AMI Labs · New York University
DES-JEPA: Learning a Shared Vocabulary for High-Dimensional Discrete Event Sequences
Oct 28, 2026Hugo Math, Maximilian Forstenhäusler, Christos AnagnostopoulosSequence Modeling Joint-Embedding Predictive Architecture
BMW Group · University of Glasgow
Self-Supervised Latent Action-Conditioned World Modeling for Time Series
Oct 27, 2026Chia-Han Kuo, Chengshuai Zhao, Huan Liu +1Self-Supervised Time Series Representation Learning Action-Conditioned World Models
Arizona State University
JEPA-Inspired Latent Field Prediction for Sparse-Pilot MIMO-OFDM Channel Reconstruction
Oct 27, 2026Yuzhi Yang, Brahim Mefgouda, Hang Zou +5Channel Estimation Channel State Prediction
Research Institute for Digital Future, Khalifa University, 127788 Abu Dhabi, UAE · School of Information Science, Japan Advanced Institute of Science and Technology, Nomi, Ishikawa, Japan · the Faculty of Applied Sciences, Macao Polytechnic University, Macao SAR, China
GUI-JEPA: Candidate-Conditioned JEPA Latents as a Fast Action Policy for Mobile GUI Agents
Oct 27, 2026Bingchao Wang, Yongcheng Jing, Shunyu Liu +5Mobile GUI Agents Mobile GUI Automation
Shanghai Jiao Tong University · Nanyang Technological University · Alibaba Cloud Computing
SWIM: Learning What Matters from Behaviors for JEPA World Models
Oct 27, 2026Jiaming Hu, Shi Bo, Yan Zheng +11World Models for Robotics World Model-Based Planning
Boston University · Unity Technologies
AC-JEPA: Anchored Cross-Modal Joint Embedding Predictive Architecture for Visual-to-Symbolic Translation
Oct 27, 2026Chaofei Hong, Ziyang Chen, Baochen Wang +2Joint-Embedding Predictive Architecture Cross-Modal Representation Learning
Zhejiang Lab
SuperJEPA: Pretraining with a Supervision-Augmented Joint-Embedding Predictive Architecture for High Energy Physics
Oct 26, 2026Dylan Sheldon Rankin, Ho Fung TsoiSelf-Supervised Pre-Training Joint-Embedding Predictive Architecture
Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, PA, USA
Fovea-JEPA: Overcoming Geometric Blind Spots in V-JEPA for Camera Motion Recognition
Oct 25, 2026Xiangkun Shi, Hanhui Li, Shiyue Zhang +4Camera Motion Understanding Video Representation Learning
Sun Yat-sen University · ByteDance
Exploring Joint-Embedding Predictive Architectures for Self-Supervised Representation Learning of Powder X-ray Diffraction Patterns
Oct 23, 2026Ming-Chiang Chang, Brian DeCostRepresentation Learning Self-Supervised Pre-Training
National Institute of Standards and Technology 100 Bureau Dr Gaithersburg, MD 20899 USA
DT-JEPA: Learning to Predict Future Representations for Decision-Making
Oct 23, 2026Changxu Ni, Yiyang Ni, Jun Li +4Transformer-Based RL Self-Supervised RL
Southeast University · Jiangsu Second Normal University · The University of Hong Kong +1
Shared Is Not Enough: Predictor Width and Factor Readability Under Masked Latent Prediction
Oct 23, 2026Nilotpal KakatiRepresentation Learning Predictive Representation Learning
Weizmann Institute of Science, Rehovot, Israel
Inside-JEPA: From Prediction to Relational Grounding in Latent World Models
Oct 22, 2026Hossein Resani, Javen Qinfeng ShiJoint-Embedding Predictive Architecture Latent World Models
Australian Institute for Machine Learning, Adelaide University · Responsible AI Research Centre
Are Joint-Embedding Predictive Objectives Effective for Longitudinal EHRs?
Oct 21, 2026Ashvin Gupta, Brendan Delaney, Alessandra RussoRepresentation Learning Joint-Embedding Predictive Architecture
Imperial College London
DepState-JEPA: Learning Depth-Grounded State Representation for Visual Perception and Manipulation
Oct 21, 2026Xiujian Liang, Liuyuan Na, Zehao Du +6State Representation Learning Visual Representation Learning
Fudan University · Shanghai Innovation Institute (SII) · Shanghai Jiao Tong University +2
OccJEPA: Occlusion as Masking for Amodal Video Segmentation
Oct 21, 2026Genpei ZhangAmodal Segmentation Video Object Segmentation
Department of Computer Sciences, University of Wisconsin–Madison
Exploring Temporal Context-Aware Physical Grounding in Latent World Models
Oct 20, 2026Yilei Man, Yang Chen, Jingcai GuoWorld Models Latent World Models
Department of Computing, The Hong Kong Polytechnic University · Department of Land Surveying and Geospatial Science, The Hong Kong Polytechnic University
LocaLeJEPA: Maximum Entropy Representations under Geometric Priors
Oct 20, 2026Kartik Tandon, Alessio Borgi, Pavel Snopov +8Representation Learning Joint-Embedding Predictive Architecture
University of Pennsylvania · Sapienza University of Rome · University of Cambridge +6
Slot-JEPA: Object-Structured Prediction Interfaces for Latent World Models
Oct 20, 2026Peisong Wen, Qianqian Xu, Minye Lei +3Goal-Conditioned Planning Joint-Embedding Predictive Architecture
School of Computer Science and Technology, University of Chinese Academy of Sciences · State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences · Beijing Academy of Artificial Intelligence +2
Predictive Closure: A Set-Level Characterization of JEPA's Feature Selection
Oct 19, 2026Mosh Levy, Aly Lidayan, Yoav GoldbergJoint-Embedding Predictive Architecture Predictive Representation Learning
Bar-Ilan University · University of California, Berkeley · Allen Institute for AI
World Model Interpretability: From Decoding Representations to Understanding Latent Worlds
Oct 18, 2026Jize Wang, Yongcheng Jing, Xinyi Le +3World Models Latent World Models
School of Automation and Intelligent Sensing, Shanghai Jiao Tong University · College of Computing and Data Science, Nanyang Technological University
M-JEPA: Predictive Self-Supervised Learning for Molecular Graphs with Scaffold-Shift Evaluation on Tox21
Oct 18, 2026Karthik Iyer, Nasser SabarGraph SSL Molecular Representation Learning
Department of Computer Science and Information Technology, La Trobe University, Melbourne, VIC 3086, Australia
MM-LeJEPA: Correlated Gaussian Regularization for Multimodal Joint-Embedding Predictive Architectures
Oct 17, 2026Haowen Bai, Rui Gong$, Meichen Liu +3Multimodal SSL Joint-Embedding Predictive Architecture
Nanyang Technological University · Hyundai Motor Group Innovation Center Singapore (HMGICS)
BeJEPA: A Goal-Conditioned Humanoid Behavior World Model via Unsupervised RL
Oct 16, 2026Liyun Yan, Haoyang Li, Yang Zhang +11Humanoid Robot Control World Models for Robotics
RL2 Lab, Shanghai Jiao Tong University · Shanghai Innovation Institute · Zhejiang University +3
EgoJEPA: Beyond Retargeting for Dexterous Manipulation from Egocentric Videos
Oct 14, 2026Meng Cao, Can Zhao, Jinjie Wei +12Robotic Manipulation Cross-Embodiment Robot Learning
Mohamed bin Zayed University of Artificial Intelligence · Shanghai Jiao Tong University · X Square Robot +2
MemWM: Associative Memory-Enhanced Latent World Models for Long-Horizon Planning
Oct 14, 2026Yuwei Sun, Siyu ZhuWorld Model Learning World Models for Robotics
Shanghai Academy of AI for Science · Fudan University
W-JEPA: Wavelet Joint-Embedding Predictive Architecture for Wearable Activity Recognition
Oct 14, 2026Yogesh Prabhu, Loki Natarajan, Jingjing ZouHuman Activity Recognition Representation Learning
Department of Computer Science and Engineering, University of California San Diego (UCSD). · Herbert Wertheim School of Public Health and Human Longevity Science, University of California San Diego (UCSD). · UCSD Shiley-Marcos Alzheimer’s Disease Research Center. +1
One-Step Prediction Fidelity Is a Poor Proxy for Planning Performance in JEPA World Models
Oct 14, 2026Joshua Cheung, Zhang HeLatent World Models World Model Evaluation
Dynamic Graph World Modelling: Learning Future Dynamics from Temporal Context
Oct 14, 2026Bo Li, Yue Yang, Yujin Huang +6Dynamic Graph Learning World Model Learning
School of Information and Communication Technology, Griffith University, Australia · Maincode, Australia · School of Computing and Information Systems, University of Melbourne, Australia +2
Learning Multimodal Sensor Fusion with JEPA Representations for Breathing Activity and Pose Recognition
Oct 11, 2026Yunxiang Luo, Constantino Álvarez Casado, Xiaoting Wu +4Multimodal Learning Multimodal Sensor Fusion
Center for Machine Vision and Signal Analysis, University of Oulu, Finland · Candour Ltd, Oulu, Finland

Oct 29, 2026J/K move · Enter open · S save

Proving and Improving the Platonic Representation Hypothesis with JEPAs

Randall Balestriero, Yann LeCun

Brown University · AMI Labs · New York University

Platonic Representation Hypothesis Cross-Modal Alignment Joint-Embedding Predictive Architecture Vision-Language Alignment Predictive Representation Learning

Open paper AskPDF arXiv

Whether models trained on different modalities converge to the same representation of the world is central to AI. The Platonic Representation Hypothesis (PRH) conjectures that they do as they scale, but without a proof it offers neither a guarantee nor a recipe. We turn the hypothesis into a theorem. First, we prove a Platonic Representation Theorem: Joint-Embedding Predictive Architectures (JEPAs) trained separately on any number of modalities learn the same representation up to one rotation whenever their positive pairs are two observations of one scene--good views--and their misalignment is bounded otherwise. Second, we show that next-token prediction is a JEPA with bad views, parts of a single document, which explains why language models align with vision only through scale. Third, we validate the theory from 145k captions to the texts of 940M web images, and uncover a scaling law of alignment that predicts unseen runs within seed noise. Good views need 42 times fewer training samples than next-token prediction on the same corpus to reach the same alignment, and a 109M-parameter text JEPA aligns with DINOv2 and CLIP's image encoders better than every language model we evaluated, up to 72B parameters, and than CLIP's text encoder: next-token prediction is a remarkably inefficient route to platonic representations. As a showcase, one rotation turns separately trained encoders into an image captioner (CIDEr .838 on COCO). Even that rotation needs no pair: on the latest pair-free benchmark, our text JEPA beats every zero-shot accuracy reported with language models up to 8B parameters, e.g., 64% against at most 56% top-5 on ImageNet-100. Positive views, and not scale alone, are thus the lever to build platonic representations.