Roz: A Taste Test for World Model Latents
cs.RO, cs.LG
Submitted: 2026-10-31
Updated: 2026-10-31
Comments: Accepted at NeurIPS 2026 Robotics World Modeling Workshop, NeurIPS 2026 Workshop CWM, and NeurIPS 2026 Workshop NEmo
Code: https://github.com/kenneth-ge/roz
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Sorting out Lipschitz function approximation
- V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning
- Motus: A Unified Latent Action World Model
- Sparse Autoencoders Find Highly Interpretable Features in Language Models
- Metrics for Finite Markov Decision Processes
- Learning Task Informed Abstractions
- DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos
- Regularisation of Neural Networks by Enforcing Lipschitz Continuity
- The Value Equivalence Principle for Model-Based Reinforcement Learning
- World Models
- Hallucination in World Models is Predictable and Preventable
- TD-MPC2: Scalable, Robust World Models for Continuous Control
- Measuring Massive Multitask Language Understanding
- Tulu 3: Pushing Frontiers in Open Language Model Post-Training
- LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels
- $\pi^{*}_{0.6}$: a VLA That Learns From Experience
- Data-Efficient Reinforcement Learning with Self-Predictive Representations
- Approximate information state for approximate planning and reinforcement learning in partially observed systems
- Wan: Open and Advanced Large-Scale Video Generative Models
- HellaSwag: Can a Machine Really Finish Your Sentence?
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation