Add WAM paper: Efficient-WAM: A 1B-Parameter World-Action Model with Low-Cost Future Imagination

[279object]

Paper

• Title: Efficient-WAM: A 1B-Parameter World-Action Model with Low-Cost Future Imagination
• Short name: Efficient-WAM
• arXiv ID: 2606.10040v2
• Paper: https://arxiv.org/pdf/2606.10040
• Authors: Jiajun Li, Tiecheng Guo, Yifan Ye, Rongyu Zhang, Xiaowei Chi, Qianpu Sun, Ying Li, Yunfan Lou, Yan Huang, Zhihe Lu, Meng Guo, Shanghang Zhang
• Published: 2026-06-08
• arXiv categories: cs.RO
• Matched keywords: WAM, world-action model, world-action models

README Entry

- **Efficient-WAM**: "Efficient-WAM: A 1B-Parameter World-Action Model with Low-Cost Future Imagination", arXiv 2026. ![](https://img.shields.io/badge/Multi--Dit-9f1239) ![](https://img.shields.io/badge/Hidden--State-fb7185)
  [[📄 Paper](https://arxiv.org/pdf/2606.10040)]

Robot Decision

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

Reason

Clearly proposes and evaluates a World-Action Model for embodied robot control using future visual prediction conditioned for action generation.

Evidence

• World-Action Model coupling future visual prediction with action generation
• experiments on RoboTwin 2.0 and real-world manipulation tasks

Taxonomy Evidence

• factorizes into video branch p(zv | o,l) and action branch p(a | o,l,s,zv)
• token-sparse video latents and asymmetric video-action denoising
• future video prediction as a compact guidance signal for action generation

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