World Action Models are Zero-shot Policies
Original research
MODEL PROFILE · World-action research
DreamZero
Jointly predicts future visual states and robot actions using a pretrained video backbone.
Research organisationNVIDIA research
Research approachJoint prediction
Published year2026
BackboneWan 2.1 (14B)
Overview
DreamZero is described in NVIDIA’s June 2026 field review as a joint-prediction world-action model. Its design adapts a Wan video backbone for planning futures and generating robot actions together.
Source record
- World Action Models are Zero-shot Policies
- NVIDIA WAM technical synthesis (Moritz Reuss, 15 Jun 2026)
Core research identity and broad method were source-checked as of 2026-11-03. Independent benchmarking, commercial availability, weights and licensing have not been assessed.