analiza aiqbench
This research introduces "temporal straightening," a technique that regularizes latent trajectories in world models to make Euclidean distance a more accurate proxy for geodesic distance during planning. By jointly learning an encoder and predictor within a Joint-Embedding Predictive Architecture (JEPA), the method stabilizes gradient-based optimization and significantly boosts success rates on goal-reaching tasks without relying on irrelevant visual features.
Temporal Straightening for Latent Planning
arXiv:2603.12231v2 Announce Type: replace Abstract: Learning good representations is essential for latent planning with world models. While pretrained visual encoders produce strong semantic visual features, they are not tailored to planning and contain information irrelevant -- or even detrimental -- to planning. Inspired by the perceptual straightening hypothesis in human visual processing, we introduce temporal straightening to improve representation learning for latent planning. Using a curvature regularizer that encourages locally straightened latent trajectories, we jointly learn an encoder and a predictor of a Joint-Embedding Predictive Architecture (JEPA) world model. We show that reducing curvature this way makes the Euclidean distance in latent space a better proxy for the geodesic distance and improves the conditioning of the planning objective. We demonstrate empirically that temporal straightening makes gradient-based planning more stable and yields significantly higher success rates across a suite of goal-reaching tasks. Our code is available at https://agenticlearning.ai/temporal-straightening.
Abstract:Learning good representations is essential for latent planning with world models. While pretrained visual encoders produce strong semantic visual features, they are not tailored to planning and contain information irrelevant -- or even detrimental -- to planning. Inspired by the perceptual straightening hypothesis in human visual processing, we introduce temporal straightening to improve representation learning for latent planning. Using a curvature regularizer that encourages locally straightened latent trajectories, we jointly learn an encoder and a predictor of a Joint-Embedding Predictive Architecture (JEPA) world model. We show that reducing curvature this way makes the Euclidean distance in latent space a better proxy for the geodesic distance and improves the conditioning of the planning objective. We demonstrate empirically that temporal straightening makes gradient-based planning more stable and yields significantly higher success rates across a suite of goal-reaching tasks. Our code is available at this https URL.
Machine Learning (cs.LG)
arXiv:2603.12231 [cs.LG]
(or arXiv:2603.12231v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2603.12231 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 236 cuvinte
Subiecte conexe
Control-Geometry Straightening for Sampling-Based Latent Planning
arXiv · 30 sept. 2026
Temporal-Distance JEPA: Plan-Aware Representation Learning for Latent World Model Predictive Control
arXiv · 29 iul. 2026
Beyond a single latent space: a dual-latent world model for long-horizon planning
arXiv · 30 sept. 2026
How Long, Not How Close: A Learned Temporal Metric for Planning in Latent World Models
arXiv · 6 oct. 2026