Robotics · 2026-10-14
GreenWAM: A World Foundation Model for Robotics
Gennady Sidorov, Valerii Kornilov, Denis Grankin, Polina Fedotova, Aleksey Postnikov
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
GreenWAM is a robotics world foundation model that can serve both world-model and policy roles across different robot bodies. It uses a shared anatomical action space for 20 robot families and joint training in visual, action-conditioned, and world–action modes, with policies formed by splitting an action expert off the generator. The authors report top-tier WorldArena 1.0 results and strong benchmark performance on RoboTwin 2.0, LIBERO(-plus), and CALVIN, plus real-humanoid adaptation and single-step distillation.
Why it matters
The authors address that robot world models and control policies are often built/evaluated separately and that robot actions are recorded in incompatible formats across bodies. GreenWAM aims to unify these roles with a shared body representation, and the paper includes controlled experiments identifying factors (robot specialization, action-expert split-off, and keeping world-model modes during adaptation) that improve both prediction and control.
Method
- Shared anatomical action space: a layout-specific codec maps each robot’s native joints into 79 anatomical slots so the same slot corresponds to the same joint motion across platforms.
- Joint video–action/world training: one model is trained on multiple modes (visual, action-conditioned, and world–action) with shared generator weights; the reasoner/VAE/generator training setup is described via robotics stage and Cosmos 3 Edge initialization.
- Policy derivation via action-expert split: for policy adaptation, an action expert is split off the generator, and policies are distilled for real-time control on a real humanoid.
Limitation
For GreenWAM on WorldArena 1.0, categories use an evaluation rule where dynamic degree, flow score and motion smoothness are capped separately against each episode’s ground truth before averaging, while competitors only provide model-level metric means; thus the recomputed aggregate differs from the official WorldArena 1.0 leaderboard score.
Abstract (from arXiv)
A robot needs a world model that predicts how its actions will change the scene and a policy that chooses those actions. Models are usually built and evaluated for only one of these roles, and every robot records its actions in its own format. We ask whether one robot world model can serve both roles across different bodies, and what makes this work. We present GreenWAM, a world foundation model for robotics, together with the recipe that produces it: (1) a shared anatomical action space that describes 20 robot families, from bimanual arms to dexterous humanoids, in common joint coordinates; (2) joint training on these families in visual, action-conditioned and world-action modes; and (3) policies that split an action expert off the generator. As a world model, GreenWAM attains top-tier trajectory, depth and 3D accuracy on WorldArena 1.0, ahead of every world-action model and video generator in our comparison. As a policy, it scores among the top entries of the RoboTwin 2.0 leaderboard, reaches the highest zero-shot LIBERO-plus success among the compared methods and is competitive with dedicated vision-language-action and world-action policies on CALVIN and LIBERO. We further adapt GreenWAM to a real humanoid, where it succeeds more often than policies adapted from $\pi_{0.5}$ and from Cosmos 3 Edge, and distill it into a single-step policy for real-time control. Controlled experiments yield three findings: robot specialization improves world prediction and control together; an action expert helps when split off at the start of adaptation; and keeping world-model modes during policy adaptation preserves robustness to scene changes.
Related papers
- The Planning Limits of Latent World Models
- D-World: An Explicit World-Language-Action Model with Recursive Self-Improvement under Dual Control
- Humanoid World Action Model With Joint State--Action Generation
- Rethinking Representations for World-Action Modeling
- OpenWAM: An Open Framework for Composable World-Action Models
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-13 · Indexed by Pub-AI: 2026-10-14 UTC
GreenWAM: A World Foundation Model for Robotics
Preprint posted: 2026-10-13 · Original record ↗
- arxiv
- 2610.18774
Explicit source versions (1)
- v1 · 2026-10-13
- Preprint available
Source date: 2026-10-13 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
A robot needs a world model that predicts how its actions will change the scene and a policy that chooses those actions. Models are usually built and evaluated for only one of these roles, and every robot records its actions in its own format. We ask whether one robot world model can serve both roles across different bodies, and what makes this work. We present GreenWAM, a world foundation model for robotics, together with the recipe that produces it: (1) a shared anatomical action space that describes 20 robot families, from bimanual arms to dexterous humanoids, in common joint coordinates; (2) joint training on these families in visual, action-conditioned and world-action modes; and (3) policies that split an action expert off the generator. As a world model, GreenWAM attains top-tier trajectory, depth and 3D accuracy on WorldArena 1.0, ahead of every world-action model and video generator in our comparison. As a policy, it scores among the top entries of the RoboTwin 2.0 leaderboard, reaches the highest zero-shot LIBERO-plus success among the compared methods and is competitive with dedicated vision-language-action and world-action policies on CALVIN and LIBERO. We further adapt GreenWAM to a real humanoid, where it succeeds more often than policies adapted from $π_{0.5}$ and from Cosmos 3 Edge, and distill it into a single-step policy for real-time control. Controlled experiments yield three findings: robot specialization improves world prediction and control together; an action expert helps when split off at the start of adaptation; and keeping world-model modes during policy adaptation preserves robustness to scene changes.