Action Jets: Learning the Local Shape of the Action-conditioned Data Manifold
cs.CV
Submitted: 2026-10-13
Updated: 2026-10-13
Comments: 34 pages, 14 figures, 23 tables. Project page: https://anton-arapin.github.io/action-jets-site
Project page: https://anton-arapin.github.io/action-jets-site
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sources
- V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning
- Verifying the Union of Manifolds Hypothesis for Image Data
- Sobolev Training for Neural Networks
- WorldScore: A Unified Evaluation Benchmark for World Generation
- CAT3D: Create Anything in 3D with Multi-View Diffusion Models
- World Models
- TD-MPC2: Scalable, Robust World Models for Continuous Control
- CameraCtrl: Enabling Camera Control for Text-to-Video Generation
- Distilling the Knowledge in a Neural Network
- Vid2World: Crafting Video Diffusion Models to Interactive World Models
- LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias
- Generalization in diffusion models arises from geometry-adaptive harmonic representations
- Diffusion Models already have a Semantic Latent Space
- Prediction, Consistency, Curvature: Representation Learning for Locally-Linear Control
- WorldModelBench: Judging Video Generation Models As World Models
- Cosmos World Foundation Model Platform for Physical AI
- Understanding the Latent Space of Diffusion Models through the Lens of Riemannian Geometry
- WorldGym: World Model as An Environment for Policy Evaluation
- Knowledge Transfer with Jacobian Matching
- Your diffusion model secretly knows the dimension of the data manifold
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models