#world-models
8 briefs tagged #world-models.
Researchers target memory gaps in embodied navigation
Two papers propose memory-based planning methods to improve visual navigation for embodied AI agents.
Researchers frame world models for robotics evaluation
Two papers outline robotics world-action models and a benchmark for long-horizon interactive model testing.
New benchmarks target robot manipulation video world models
H2R-Bench and DreamX-Phi 1.0 focus on cross-embodiment and action-conditioned robot video generation.
DreamX-Phi 1.0 targets robotic manipulation rollouts
The action-conditioned video world model predicts future robot observations from frames, instructions and actions.
Alaya-EVOKE targets persistent interactive world models
The paper proposes external world-state memory and long-horizon supervision for interactive generation.
New papers target scalable training for LLM agents
EnvACE and State2State reduce reliance on manual environments, expert trajectories and task design.
Researchers propose MASS for multiplayer world models
The system separates shared world state from rendering to improve multi-agent simulation consistency.
WorldCycle targets drift in video world models
The arXiv paper proposes self-verifiable RL using reversible action cycles.