ZeroHour
arXiv cs.AI / cs.LG / cs.CLpublished ()ingested Taoyong Cui
Part of a story covered by 2 sources: “JEPA-Anything: Learning Predictive Models across Different Worlds” — merged summary and timeline →

JEPA-Anything: Learning Predictive Models across Different Worlds

infoAI researchimportance 35
AI summary · glm-5.3-flash

Researchers introduce JEPA-Anything, a domain-agnostic predictive world-modeling framework using orthogonal predictive factorization, validated across seven domains including biology and weather.

JEPA-Anything extends joint-embedding predictive architectures through orthogonal predictive factorization (OPF), which decomposes latent targets into complementary factors learned via dedicated pathways. It is evaluated across seven domains: vision, biology, clinical trajectories, control, molecular dynamics, physical fields, and weather. Against matched JEPA baselines it improves metrics on all 10 dynamics tasks, cuts Interventional Pong single-intervention error by 34.8%, and achieves the lowest one-step and 100-step molecular errors across four systems. A factor-nominated biological intervention received experimental support in cell co-cultures, patient-derived organoids, tumor fragments, and mice.

  • Introduces orthogonal predictive factorization extending JEPA to heterogeneous domains
  • Evaluated on vision, biology, clinical, control, molecular, field, and weather tasks
  • Reduces Interventional Pong error 34.8% and wins all 10 dynamics tasks
  • Latent orbital modes recover Keplerian scaling exponent (slope -1.4991)
  • Biological intervention validated in organoids, tumor fragments, and mice
OrganizationsGen-Verse
AI modelsJEPA-Anything
Full article206 words · extracted from arxiv.org · click to collapse

World modeling enables intelligence to anticipate consequences, guide interventions, and learn from interaction. Yet predictive models remain domain-specific: can a common learning principle support world modeling across radically different systems? We introduce JEPA-Anything, a domain-agnostic framework based on orthogonal predictive factorization (OPF). Extending joint-embedding predictive architectures, OPF decomposes latent targets into complementary factors, learns them through dedicated pathways, and recombines them within a shared predictive design. We evaluate JEPA-Anything across seven domains: vision, biology, clinical trajectories, control, molecular dynamics, physical fields, and weather. Experiments span representation learning, intervention prediction, out-of-distribution generalization, and long-horizon dynamics, including 10 matched dynamics tasks, forecasting of over 1,000 clinical events, and 100-step molecular rollouts across four systems. Against matched JEPA baselines, JEPA-Anything improves reported metrics on all 10 dynamics tasks and reduces single-intervention prediction error on Interventional Pong by 34.8%. It achieves the lowest one-step and 100-step molecular errors among compared methods in all four systems. Beyond prediction, a factor-nominated biological intervention receives experimental support in cell co-cultures, patient-derived organoids, tumor fragments, and mice; latent orbital modes recover the Keplerian scaling exponent with a fitted slope of -1.4991. These results support a common factorized predictive principle across heterogeneous worlds, connecting world modeling with intervention and experimentally grounded scientific discovery. Code: https://github.com/Gen-Verse/JEPA-Anything

Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.20800