New paper challenges JEPA feature-selection theory: unpredictable features can be kept
andreea7b · x · 2026-10-29
A new paper on JEPA world models revisits which features they keep. Earlier work suggested JEPA would drop features it cannot predict individually, like a traffic light. The authors present evidence this is wrong and introduce a new characterization explaining how world models actually select and retain features as abstractions for decision making.
More from Research
- New AI security benchmarks from Meta collaborators urge community to double down on robustness — maksym_andr · 2026-10-29
- Arena Releases Alignment Index, Arguing Alignment Is a Function of Model Capability — arena · 2026-10-29
- Researchers Push Back on AI Audit of Saez-Zucman Papers Over Missing Prompts — felpix_ · 2026-10-29
- Sitting between two data vendors: one sells the benchmark, one sells the hill-climbing data — xeophon · 2026-10-29
- SSIO: self-supervised inertial odometry trains across pedestrians, drones, quadrupeds and humanoids without labels — GuanyaShi · 2026-10-29
- SREGym to debut at Bench Summit, targeting AI SRE evaluation frontier — tianyin_xu · 2026-10-29