Low-Rank Dynamics-Effective Latent Carriers for Counterfactual Rollout in Learned World Models
cs.RO, cs.AI
Submitted: 2026-08-15
Updated: 2026-09-07
Comments: Revised manuscript with expanded Joint-intervention, carrier-relative, and recurrent-dynamics analyses. 54 pages, 7 figures. Code and data are available at https://github.com/lysea8282/dynamic-effective-latent-carriers
Code: https://github.com/lysea8282/dynamic-effective-latent-carriers
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Causal Physics Steering in Video World Models via Concept Activation Vectors
- V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning
- Physion: Evaluating Physical Prediction from Vision in Humans and Machines
- CRONOS: Benchmarking Counterfactual Physical Consistency in Video Models
- IntPhys 2: Benchmarking Intuitive Physics Understanding In Complex Synthetic Environments
- Hilbert's sixth problem: derivation of fluid equations via Boltzmann's kinetic theory
- Intuitive physics understanding emerges from self-supervised pretraining on natural videos
- World Models
- Mastering Diverse Domains through World Models
- Interpreting Physics in Video World Models
- Latent State Design for World Models under Sufficiency Constraints
- IntPhys: A Framework and Benchmark for Visual Intuitive Physics Reasoning
- Physically Viable World Models: A Case for Query-Conditioned Embodied AI
- Steering Language Models With Activation Engineering
- GAUGE: A Measurement-Grounded Benchmark for Physical Fidelity in Simulation Engines and Video World Models
- SlotFormer: Unsupervised Visual Dynamics Simulation with Object-Centric Models
- Is Forward Prediction Enough? Physical State Grounding for JEPA World Models
- What Makes Video World Model Latents Action-Relevant: Prediction over Reconstruction
- CLEVRER: CoLlision Events for Video REpresentation and Reasoning
- What Do World Models Learn in RL? Probing Latent Representations in Learned Environment Simulators
Related papers
- FMT x: An Efficient and Asymptotically Optimal Extension of the Fast Marching Tree for Dynamic Replanning
- MPCFormer: A physics-informed data-driven approach for explainable socially-aware autonomous driving
- RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies
- HRDexDB: A 4D Dexterous Grasping Dataset Across Human and Multiple Robot Embodiments
- APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies
- Fine-tuning is Not Enough: A Parallel Framework for Collaborative Imitation and Reinforcement Learning in End-to-end Autonomous Driving