Add WAM paper: EgoWAM: World Action Models Beyond Pixels with In-the-Wild Egocentric Human Data

[279object]

Paper

• Title: EgoWAM: World Action Models Beyond Pixels with In-the-Wild Egocentric Human Data
• Short name: EgoWAM
• arXiv ID: 2607.08436v1
• Paper: https://arxiv.org/pdf/2607.08436
• Authors: Baoyu Li, Xinchen Yin, Mengying Lin, Yixin Zhang, Danfei Xu
• Published: 2026-07-08
• arXiv categories: cs.RO, cs.AI
• Matched keywords: WAM, world action model, world action models

README Entry

- **EgoWAM**: "EgoWAM: World Action Models Beyond Pixels with In-the-Wild Egocentric Human Data", arXiv 2026. ![](https://img.shields.io/badge/Multi--Dit-9f1239) ![](https://img.shields.io/badge/Shared--Rep-db2777)
  [[📄 Paper](https://arxiv.org/pdf/2607.08436)]

Robot Decision

• Decision: accept
• Confidence: high
• Taxonomy: joint_diffusion_multistream_shared_representation
• Taxonomy confidence: medium
• README heading: ### Diffusion-based Generation
• Badges: https://img.shields.io/badge/Multi--Dit-9f1239, https://img.shields.io/badge/Shared--Rep-db2777

Reason

The paper explicitly introduces and evaluates World Action Models for robot manipulation, using action-conditioned future scene/world prediction as auxiliary supervision for policies.

Evidence

• World Action Models (WAMs) require policies to predict actions and how the scene evolves
• robot manipulation co-training with action head and world-model head
• compares world prediction targets: Pixel, DINO, and 3D motion flow

Taxonomy Evidence

• shared encoder with action head and world-model head
• jointly predicts actions and future world targets Pixel, DINO, and 3D flow
• co-trains on lab robot data and in-the-wild egocentric human data

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