RESEARCH WATCH ARCHIVE
Research papers
Browse 53 records. Automated additions contain bibliographic metadata only, with no claim of editorial verification.
History Forcing: Distilling History-Dependent Continuations for Long-Horizon Video World Models
Right Ranking, Wrong Scale: Reproducing the Robotic World Model's Dynamics Model and Testing the Uncertainty of Its Follow-up
Visuo-tactile Diffusion Policy for Robotic Manipulation with Auxiliary Tactile Dynamics Prediction
Gnomon4D: Learning Where and When to Focus for Geometric Motion-Grounded World-Action Modeling
AgiWorld: An Agentic Video World Model for Building Persistent Worlds
OpenNWM: A Generalist Navigation World Model with Latent Action Pretraining
When Does Latent Prediction Error Support Physical Surprise Claims in Video World Models?
Xiaomi Auto World Model 2.0: Causal World Models for Real-World Driving and Embodied Intelligence
Stand-In Effects and Pooling Amplification: Design-Based Tests of Whether Biological World Models Use Actions
Vision-Language Models Are Joint Denoisers in Robotic Action Generation
pico-WAM: Evaluating Counterfactual Prediction and Planning Error in a Compact World Action Model
Why Koopman Structure Does Not Help Differentiable World Models for Humanoid Whole-Body Tracking
Old Is Not False: Valid-Time Memory for Interactive Video World Models
WorldSketch: Sketching World Models for Zero-Shot Mobile Manipulation
AuroraWorld: Structured 4D World Modeling for Exo-to-Ego Video Generation in Robot Manipulation
SparseWAM: Driving World-Action Models are Worth 16 Register Tokens
FORCE: Factorized Object Relations for Counterfactual Editing in Video World Models
RoboSeed: Multidimensional Evaluation of Vision-Language-Action and World-Action Models: Capability, Resource Affordability, and Research Evidence
TurboWAM: Learning What Matters for Fast World Action Models
H-JEPA: End-to-End Learning of Hierarchical World Models for Visual Planning
SV-WAM: An Efficient Surround-View World-Action Model for End-to-End Autonomous Driving
Robot-Factored World Models via Robot Rendering
ReWorld: Multi-Dimensional Reward Modeling for Embodied Video Generation
From Prediction to Control: A Survey of Closed-Loop World Models for Robot Learning
WorldBridge: Bridging Video and Action for Task Generalization in World Action Models
Explores how new-task video supervision and action grounding interact.
What Does World Modeling Actually Improve in Robot Action Learning?
Controlled study of predictive supervision in robot action learning; results under distribution shift are particularly relevant.
Agentic World Action Models
Research combining reusable code-based skills with video and action prediction.
Plan and Double-Check: Streaming Multimodal Q-Former for Online Robot Action Generation
World-Action Models for Robot Learning and Control: A Survey
A source-linked survey of world-action architectures, data, benchmarks and research challenges.
World Model for Robot Learning: A Comprehensive Survey
A broader synthesis of world models in robot planning, simulation, evaluation and learning.
Being-H0.7: A Latent World-Action Model from Egocentric Videos
Future-informed latent policy learning without runtime visual rollouts.
Fast-WAM: Do World Action Models Need Test-time Future Imagination?
Original Fast-WAM research about test-time video generation.
V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning
Dense visual and temporal representations relevant to predictive control and robot perception.
World Action Models are Zero-shot Policies
Original DreamZero model research.
BagelVLA: Enhancing Long-Horizon Manipulation via Interleaved Vision-Language-Action Generation
A predictive vision-language-action architecture for multi-stage tasks.
DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos
Latent-action conditioning for controllable world models.
Causal World Modeling for Robot Control
Original LingBot-VA robot control research.
Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control and Planning
Original Cosmos Policy research.
mimic-video: Video-Action Models for Generalizable Robot Control Beyond VLAs
Video-model latent plans decoded into robot actions.
Motus: A Unified Latent Action World Model
A mixture-of-experts approach spanning prediction and control.
Unified World Models: Coupling Video and Action Diffusion for Pretraining on Large Robotic Datasets
V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning
Video-based representation learning and action-conditioned planning with robot interaction data.
Unified Video Action Model
Shared video-action representation with separate inference pathways.
Cosmos World Foundation Model Platform for Physical AI
A video world-foundation-model platform aimed at physical-AI development and simulation.
Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representations
Video prediction as a way to learn action-relevant visual features.
Generative Image as Action Models
Image-space action generation; adjacent to video-based policy learning.
DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset
Robot-interaction datasets can support training and evaluation of action-generating systems.
Genie: Generative Interactive Environments
Controllable generated environments offer a related direction for learning world dynamics.
Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulation
Video-pretrained joint prediction of images and robot actions.
Diffusion Policy: Visuomotor Policy Learning via Action Diffusion
Action generation through diffusion is an important adjacent approach to visuomotor control.
UniPi: Learning universal policies via text-guided video generation
Original UniPi text-guided video policy research.