H-JEPA paper separates perception and control, hits 91.9% on OGB-Cube in 10 epochs
ylecun · x · 2026-10-04
An arXiv paper by Tamim Zoabi, Ameen Ali, and Lior Wolf — retweeted by Yann LeCun — introduces H-JEPA, an action-conditioned world model that splits perception from control. A wide perceptual code is regularized toward an isotropic geometry with a Bures-Wasserstein prior, and a fixed orthonormal slice serves as the control state, evolving under phase-conditioned dissipative port-Hamiltonian dynamics. Port-inverse consistency (PIC) reads actions back through the port transpose and is provably a parameter-free reweighting of rollout error. H-JEPA matches or beats reconstruction-free baselines including Delta-JEPA on four pixel-based control benchmarks within 10 training epochs, with its largest gain on OGB-Cube (91.9% vs 79.3%). Ablations show untying the readout from the port halves the gain.
More from Research
- Matt Green muses: are n^25 galactic algorithms normal, or a glimpse of provable Algorithmica? — matthew_d_green · 2026-11-20
- MIT and NVIDIA's QDA uses agents to auto-design LLM quantization recipes, beating top public ones by up to 2.2 points — omarsar0 · 2026-11-20
- AI-powered forensic metascience: auditing 7 million papers, 500K files scanned, 100+ issues found — gleech · 2026-11-20
- Harmonic's Aristotle writes proven-correct fusion plasma controllers in C and Verilog in 4 hours — satnam6502 · 2026-11-20
- Interp4discovery workshop on interpretability for scientific discovery at NeurIPS, Dec 12 — begusgasper · 2026-11-20
- Talk compares tabular foundation models TabPFN, TabICR vs gradient boosting — saheedniyi_02 · 2026-11-20