Causal-JEPA: Learning World Models through Object-Level Latent Masking

Heejeong Nam, Quentin Le Lidec, Lucas Maes, Yann Lecun, Randall Balestriero
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:91808-91832, 2026.

Abstract

World models require robust relational understanding to support prediction, reasoning, and control. While object-centric representations provide a useful abstraction, they are not sufficient to capture interaction-dependent dynamics. We therefore propose C-JEPA, a simple and flexible object-centric world model that extends masked joint embedding prediction from image patches to object-centric representations. By masking object-level latents and requiring each masked object state to be inferred from the surrounding context, C-JEPA imposes structured partial observability during training, creating counterfactual-like prediction queries that discourage shortcut solutions and make interaction-dependent prediction necessary under the learning objective. Empirically, C-JEPA leads to consistent gains in visual question answering, with an absolute improvement of about 20% in counterfactual reasoning over the same architecture without object-level masking. On agent control tasks, C-JEPA enables substantially more efficient planning by using only 1% of the total latent input features required by patch-based world models, while achieving comparable performance. Finally, we provide a formal analysis demonstrating that object-level masking induces useful inductive bias by controlling observability. Our code is available at https://github.com/galilai-group/cjepa.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-nam26c, title = {Causal-{JEPA}: Learning World Models through Object-Level Latent Masking}, author = {Nam, Heejeong and Lidec, Quentin Le and Maes, Lucas and Lecun, Yann and Balestriero, Randall}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {91808--91832}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/nam26c/nam26c.pdf}, url = {https://proceedings.mlr.press/v306/nam26c.html}, abstract = {World models require robust relational understanding to support prediction, reasoning, and control. While object-centric representations provide a useful abstraction, they are not sufficient to capture interaction-dependent dynamics. We therefore propose C-JEPA, a simple and flexible object-centric world model that extends masked joint embedding prediction from image patches to object-centric representations. By masking object-level latents and requiring each masked object state to be inferred from the surrounding context, C-JEPA imposes structured partial observability during training, creating counterfactual-like prediction queries that discourage shortcut solutions and make interaction-dependent prediction necessary under the learning objective. Empirically, C-JEPA leads to consistent gains in visual question answering, with an absolute improvement of about 20% in counterfactual reasoning over the same architecture without object-level masking. On agent control tasks, C-JEPA enables substantially more efficient planning by using only 1% of the total latent input features required by patch-based world models, while achieving comparable performance. Finally, we provide a formal analysis demonstrating that object-level masking induces useful inductive bias by controlling observability. Our code is available at https://github.com/galilai-group/cjepa.} }
Endnote
%0 Conference Paper %T Causal-JEPA: Learning World Models through Object-Level Latent Masking %A Heejeong Nam %A Quentin Le Lidec %A Lucas Maes %A Yann Lecun %A Randall Balestriero %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-nam26c %I PMLR %P 91808--91832 %U https://proceedings.mlr.press/v306/nam26c.html %V 306 %X World models require robust relational understanding to support prediction, reasoning, and control. While object-centric representations provide a useful abstraction, they are not sufficient to capture interaction-dependent dynamics. We therefore propose C-JEPA, a simple and flexible object-centric world model that extends masked joint embedding prediction from image patches to object-centric representations. By masking object-level latents and requiring each masked object state to be inferred from the surrounding context, C-JEPA imposes structured partial observability during training, creating counterfactual-like prediction queries that discourage shortcut solutions and make interaction-dependent prediction necessary under the learning objective. Empirically, C-JEPA leads to consistent gains in visual question answering, with an absolute improvement of about 20% in counterfactual reasoning over the same architecture without object-level masking. On agent control tasks, C-JEPA enables substantially more efficient planning by using only 1% of the total latent input features required by patch-based world models, while achieving comparable performance. Finally, we provide a formal analysis demonstrating that object-level masking induces useful inductive bias by controlling observability. Our code is available at https://github.com/galilai-group/cjepa.
APA
Nam, H., Lidec, Q.L., Maes, L., Lecun, Y. & Balestriero, R.. (2026). Causal-JEPA: Learning World Models through Object-Level Latent Masking. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:91808-91832 Available from https://proceedings.mlr.press/v306/nam26c.html.

Related Material