JEPA-Anything uses one recipe across seven fields, beating baselines on dynamics tasks
Original titleBeyond Domain-Specific World Models: JEPA-Anything Uses 1 Recipe for 7 Fields
- Outlet
- MarkTechPost
- Published
- Importance
- 3/5
Summary
JEPA-Anything splits a JEPA's single latent target into four orthogonal factors, each with its own predictor. Tested across seven domains, it outperformed matched JEPA baselines on all ten dynamics tasks and reduced Interventional Pong intervention error by 34.8 percent.
Original excerpt
JEPA-Anything splits a JEPA's single latent target into 4 orthogonal factors, each with its own predictor. Tested across 7 domains, it beat matched JEPA baselines on all 10 dynamics tasks and cut Interventional Pong intervention error by 34.8%. The post Beyond Domain-Specific World Models: JEPA-Anything Uses 1 Recipe for 7 Fields appeared first on MarkTechPost.
#jepa#world-model#representation-learning