# EgoGenesis: Egocentric World-Action Modeling with Online Anchored Projective Memory and Action-3D RoPE

arXiv: https://arxiv.org/abs/2607.28243 · HTML: https://mlsanity.com/paper/2607.28243/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Embodied and robot world models, one of 11 directions in World models.
- Builds on 17 of its 51 references that are in the map.
- First posted on arXiv 30 Jul 2026.

## Placed in

- World models / [Embodied and robot world models](https://mlsanity.com/wm/embodied-and-robot-world-models/)

## Builds on (most cited first, 17 of 17)

- [Wan: Open and Advanced Large-Scale Video Generative Models](https://mlsanity.com/paper/2503.20314/) (2025)
- [AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems](https://mlsanity.com/paper/2503.06669/) (2025)
- [Causal World Modeling for Robot Control](https://mlsanity.com/paper/2601.21998/) (2026)
- [EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric Video](https://mlsanity.com/paper/2505.11709/) (2025)
- [Ctrl-World: A Controllable Generative World Model for Robot Manipulation](https://mlsanity.com/paper/2510.10125/) (2025)
- [Cosmos 3: Omnimodal World Models for Physical AI](https://mlsanity.com/paper/2606.02800/) (2026)
- [RoboTwin: Dual-Arm Robot Benchmark with Generative Digital Twins](https://mlsanity.com/paper/2504.13059/) (2025)
- [VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Model](https://mlsanity.com/paper/2602.10098/) (2026)
- [Cameras as Relative Positional Encoding](https://mlsanity.com/paper/2507.10496/) (2025)
- [ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?](https://mlsanity.com/paper/2606.19531/) (2026)
- [EgoWAM: World Action Models Beyond Pixels with In-the-Wild Egocentric Human Data](https://mlsanity.com/paper/2607.08436/) (2026)
- [EgoSim: Egocentric World Simulator for Embodied Interaction Generation](https://mlsanity.com/paper/2604.01001/) (2026)
- [InterDyn: Controllable Interactive Dynamics with Video Diffusion Models](https://mlsanity.com/paper/2412.11785/) (2024)
- [Egocentric World Model for Photorealistic Hand-Object Interaction Synthesis](https://mlsanity.com/paper/2603.13615/) (2026)
- [EgoInfinity: A Web-Scale 4D Hand-Object Interaction Data Engine for Any-View Robot Retargeting and Video-to-Action Robot Learning](https://mlsanity.com/paper/2606.17385/) (2026)
- [Masked Visual Actions for Unified World Modeling](https://mlsanity.com/paper/2607.19343/) (2026)
- [RynnWorld-Teleop: An Action-Conditioned World Model for Digital Teleoperation](https://mlsanity.com/paper/2607.06558/) (2026)

## Built on by (most cited first, 2 of 2)

- [Acting in Meters: Learning Metric Interactions for Precise Robotic Manipulation](https://mlsanity.com/paper/2609.18243/) (2026)
- [Multi-Agent Egocentric World Model with Fine-Grained Embodied Interaction](https://mlsanity.com/paper/2610.12299/) (2026)

## Abstract (arXiv)

Egocentric video offers rich manipulation experience for embodied AI, yet collecting diverse egocentric data across scenes, objects, motions, and embodiments remains costly. We present method, an egocentric world-action simulator that synthesizes controllable, high-quality manipulation videos to expand scarce real-world training data. method builds on a pretrained video generation prior and introduces two geometry-aware conditioning mechanisms. Online Anchored Projective Memory (OAPM) preserves a first-frame 3D scene anchor while periodically refreshing a recent state during autoregressive generation. Action-3D Rotary Position Embedding (A3D-RoPE) encodes end-effector motion with camera-aware 3D rotary coordinates, injecting action geometry into skeleton-to-video cross-attention for precise control. Together, these components improve visual fidelity, geometric stability, and action alignment in long egocentric rollouts. Moreover, augmenting 400 real trajectories with 400 method-generated trajectories improves out-of-distribution real-robot success from 77% to 84% on single-arm tasks and from 53% to 70% on dual-arm tasks, demonstrating that the synthesized data substantially improve downstream WAM generalization.