aiqbench analysis
This study presents a training recipe allowing Joint-Embedding Predictive Architectures (JEPA) to function alongside value objectives in tabular foundation models, preventing the latent term from collapsing. While the approach successfully preserves world-model capabilities, empirical results show it incurs higher computational costs and yields mixed performance gains compared to standard value-only baselines across 147 datasets. The work offers a practical framework for researchers exploring self-supervised learning in structured data without overpromising immediate efficiency or accuracy breakthroughs.
A JEPA Recipe for Tabular Foundation Models
arXiv:2609.25541v1 Announce Type: cross Abstract: Tabular foundation models learn to predict cell values in context, whereas world-model self-supervision asks for prediction in representation space (LeCun, 2022; Assran et al., 2023). On a tabular foundation-model prior, the latent term of a joint-embedding predictive architecture (JEPA) collapsed in our earlier runs and took the encoder with it to a constant map. We report a recipe under which the latent term survives to convergence beside the value objective: the value head reads the encoder field rather than the predictor, and the target is an exponential moving average (EMA) difference. To bound its cost against the value-only arm, both arms train until a plateau rule stops them, with no fixed step budget. A fixed horizon had confounded a slowdown with a ceiling, since the value-only arm was still improving well past the usual budget. At convergence, in one run per arm, the JEPA arm trails the value-only arm across 147 real datasets, 32:70 wins to losses on classification (29:63 with one entry per dataset name) and 8:24 on regression, the margin small on classification and wider on regression, and the count leans the same way in each stratum and each benchmark. The JEPA arm (jepa) needs 1.42 times as many steps as the value-only arm (ds), and 1.66 times its wall-clock, to reach its plateau.
Abstract:Tabular foundation models learn to predict cell values in context, whereas world-model self-supervision asks for prediction in representation space (LeCun, 2022; Assran et al., 2023). On a tabular foundation-model prior, the latent term of a joint-embedding predictive architecture (JEPA) collapsed in our earlier runs and took the encoder with it to a constant map. We report a recipe under which the latent term survives to convergence beside the value objective: the value head reads the encoder field rather than the predictor, and the target is an exponential moving average (EMA) difference. To bound its cost against the value-only arm, both arms train until a plateau rule stops them, with no fixed step budget. A fixed horizon had confounded a slowdown with a ceiling, since the value-only arm was still improving well past the usual budget. At convergence, in one run per arm, the JEPA arm trails the value-only arm across 147 real datasets, 32:70 wins to losses on classification (29:63 with one entry per dataset name) and 8:24 on regression, the margin small on classification and wider on regression, and the count leans the same way in each stratum and each benchmark. The JEPA arm (jepa) needs 1.42 times as many steps as the value-only arm (ds), and 1.66 times its wall-clock, to reach its plateau.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2609.25541 [cs.LG]
(or arXiv:2609.25541v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.25541 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 313 words
Related coverage
NVIDIA Kumo Tabular Sets a New Accuracy-Efficiency Frontier for Tabular Prediction
Hugging Face · Sep 29, 2026
LeRobot v0.6.0: Imagine, Evaluate, Improve
Hugging Face · Jul 7, 2026
Recurrent In-Context Learning for Tabular Foundation Models
arXiv · Oct 15, 2026
Introducing TabFM: A zero-shot foundation model for tabular data
Google Research · Jun 30, 2026