Add WAM paper: MotionWAM: Towards Foundation World Action Models for Real-Time Humanoid Loco-Manipulation

[279object]

Paper

• Title: MotionWAM: Towards Foundation World Action Models for Real-Time Humanoid Loco-Manipulation
• Short name: MotionWAM
• arXiv ID: 2606.09215v1
• Paper: https://arxiv.org/pdf/2606.09215
• Authors: Jia Zheng, Teli Ma, Yudong Fan, Zifan Wang, Shuo Yang, Junwei Liang
• Published: 2026-06-08
• arXiv categories: cs.RO
• Matched keywords: WAM, world action model, world action models

README Entry

- **MotionWAM**: "MotionWAM: Towards Foundation World Action Models for Real-Time Humanoid Loco-Manipulation", arXiv 2026. ![](https://img.shields.io/badge/Multi--Dit-9f1239) ![](https://img.shields.io/badge/Hidden--State-fb7185)
  [[📄 Paper](https://arxiv.org/pdf/2606.09215)]

Robot Decision

• Decision: accept
• Confidence: high
• Taxonomy: joint_diffusion_multistream_hidden_state
• Taxonomy confidence: high
• README heading: ### Diffusion-based Generation
• Badges: https://img.shields.io/badge/Multi--Dit-9f1239, https://img.shields.io/badge/Hidden--State-fb7185

Reason

Clearly presents and evaluates a World Action Model for action-conditioned humanoid loco-manipulation control.

Evidence

• Title explicitly says Foundation World Action Models
• conditions policy on video world model features to predict whole-body motion tokens
• real-world humanoid loco-manipulation and control evaluation

Taxonomy Evidence

• policy conditioned on intermediate denoising features of a video world model
• Video DiT with Motion DiT through intermediate denoising features
• Motion DiT consumes video hidden states to output whole-body motion tokens

Human Review Checklist

• This paper belongs in the WAM survey
• README section is correct
• Badges are correct
• Short name is correct
• Paper link is correct
• Merge this PR if accepted