PolyU OSWAM distills World Action Models to one denoising step for real-time control
PolyUHK · hf · 2026-11-02
Hong Kong Polytechnic University's VCLab released OSWAM, a World Action Model achieving one-step action generation for autonomous driving and embodied control. Direct-policy WAMs still need tens of denoising steps for competitive action quality; naive step reduction collapses the trajectory. OSWAM initializes diffusion with structured imagination tokens and distills future-conditioned privileged information into a single denoising step, enabling real-time, high-quality action generation. Code and project page are available.
More from Embodied
- Ant Group's Robbyant shows LingBot-VLA 2.0: 60k hours of real data, runs on 20+ robots — huangyun_122 · 2026-11-03
- Ropedia W0: cross-embodiment world model trained on 10k hours of human video hits SOTA — liuziwei7 · 2026-11-03
- Delta Intelligence unveils Δ₀, a humanoid foundation model for whole-body loco-manipulation — chris_j_paxton · 2026-11-03
- Phoenix Tech Report Tackles Why Robots Fail at Final Contact — jiqizhixin · 2026-11-03
- Tech Sales Veteran Reviews iPhone Duo: Worth It, But App Support Lags — thedealdirector · 2026-11-03
- Tutor Intelligence raises $75M Series B led by a16z to deploy teleop-trained robots — a16z · 2026-11-03