Add WAM paper: GigaWorld-Policy-0.5: A Faster and Stronger WAM Empowered by AutoResearch

[279object]

Paper

• Title: GigaWorld-Policy-0.5: A Faster and Stronger WAM Empowered by AutoResearch
• Short name: GigaWorld-Policy-0.5
• arXiv ID: 2607.13960v1
• Paper: https://arxiv.org/pdf/2607.13960
• Authors: GigaWorld Team, Angen Ye, Angyuan Ma, Boyuan Wang, Chaojun Ni, Fangzheng Ye, Guan Huang, Guo Li, Guosheng Zhao, Haodong Yan, Hengtao Li, Jiwen Lu, Kai Wang, Mingming Yu, Qitang Hu, Qiuping Deng, Songling Liu, Xiaoyu Tian, Xiaofeng Wang, Xinyu Zhou, Xiuwei Xu, Xinze Chen, Yang Wang, Yejun Zeng, Yifan Chang, Yun Ye, Zhenyu Wu, Zhanqian Wu, Zheng Zhu
• Published: 2026-07-15
• arXiv categories: cs.RO
• Matched keywords: WAM, world action model, world action models

README Entry

- **GigaWorld-Policy-0.5**: "GigaWorld-Policy-0.5: A Faster and Stronger WAM Empowered by AutoResearch", arXiv 2026. ![](https://img.shields.io/badge/Unified--DiT-be123c) ![](https://img.shields.io/badge/Explicit-f43f5e)
  [[📄 Paper](https://arxiv.org/pdf/2607.13960)]

Robot Decision

• Decision: accept
• Confidence: high
• Taxonomy: joint_diffusion_unified_explicit
• Taxonomy confidence: medium
• README heading: ### Diffusion-based Generation
• Badges: https://img.shields.io/badge/Unified--DiT-be123c, https://img.shields.io/badge/Explicit-f43f5e

Reason

The paper explicitly presents an action-centered World Action Model for robot control, jointly modeling actions and future visual dynamics with action-only inference.

Evidence

• explicitly defines GigaWorld-Policy-0.5 as an enhanced action-centered WAM
• jointly predicts action chunks and future visual observations for robot policy learning
• uses Action-Conditioned World Modeling and WAM training for robot control

Taxonomy Evidence

• jointly modeling actions and future visual observations
• future visual dynamics are used during training while action-only decoding is used at inference time
• future visual tokens are predicted conditioned on the action sequence
• Mixture-of-Transformers architecture separates visual dynamics modeling and action generation into specialized experts

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