@randall_balestr: Delighted to be speaking about our recent progress on JEPAs and on the need for more principled solutions that handle d…
Summary
Randal Balestrero announces his keynote talk at MICCAI-FOMO26 on recent progress in Joint-Embedding Predictive Architectures (JEPAs) and the need for more principled approaches to data imbalance, noise, multimodality, and reasoning.
View Cached Full Text
Cached at: 10/01/26, 04:26 PM
Delighted to be speaking about our recent progress on JEPAs and on the need for more principled solutions that handle data imbalance, noise, multimodality, reasoning, …! Happening at MICCAI-FOMO26 in 5min! And thank you for having me as keynote today! https://t.co/nN98CNZF8C
Similar Articles
@randall_balestr: Will be speaking at Harvard today at 9:45 about world models, Le* family and why we need more theory and mathematics to…
The author announces a speaking engagement at Harvard University where they will discuss world models, the Le* family, and the need for more theory and mathematics to advance JEPAs.
@AbdelStark: It’s time to JEPA pill the world! awesome-jepa: A curated list of papers, models, code, datasets, and learning resource…
A curated list of papers, models, code, datasets, and learning resources for Joint Embedding Predictive Architectures (JEPA), the self-supervised approach to world models proposed by Yann LeCun.
@BasileTerv987: Accepted to TMLR, with reproducibility certification v2 of our JEPA-WM study (arXiv:2512.24497) is out, with new data-s…
Basile Terver and colleagues' paper on Joint-Embedding Predictive World Models (JEPA-WM) for robotics has been accepted to TMLR with a reproducibility certification. The updated version includes new data-scaling experiments, a Lipschitz analysis of multistep rollout training, and extended discussions.
Representation Without Reward: A JEPA Audit for LLM Fine-Tuning
This paper audits Joint-embedding predictive architectures (JEPA) for LLM fine-tuning on a natural-language-to-regex task, testing twenty-two auxiliary objectives. The results show that hidden-state representation improvements are only weakly coupled to decoded-task accuracy, with no auxiliary surviving family-wise correction.
@jaminball: I love attending conferences focused on research, and @CAISconf this week was great! Hearing what's happening on the bl…
Percy Liang discussed training open frontier models with modest compute, emphasizing algorithmic efficiency and scaling recipes, and advocated for open development beyond open weights.