LDA-1B: Scaling Latent Dynamics Action Model via Universal Embodied Data Ingestion

TL;DR

LDA-1B is a large-scale robot foundation model that ingests diverse embodied data to learn dynamics, policy, and visual forecasting, enabling improved performance and data efficiency in robotic tasks.

Contribution

The paper introduces LDA-1B, a scalable robot foundation model that leverages heterogeneous embodied data through structured latent space prediction and multi-modal transformers.

Findings

Outperforms prior methods by up to 21%, 48%, and 23% on various tasks.

Assembles and standardizes a large dataset of 30k hours of embodied interactions.

Enables data-efficient fine-tuning, improving performance with low-quality data.

Abstract

Recent robot foundation models largely rely on large-scale behavior cloning, which imitates expert actions but discards transferable dynamics knowledge embedded in heterogeneous embodied data. While the Unified World Model (UWM) formulation has the potential to leverage such diverse data, existing instantiations struggle to scale to foundation-level due to coarse data usage and fragmented datasets. We introduce LDA-1B, a robot foundation model that scales through universal embodied data ingestion by jointly learning dynamics, policy, and visual forecasting, assigning distinct roles to data of varying quality. To support this regime at scale, we assemble and standardize EI-30k, an embodied interaction dataset comprising over 30k hours of human and robot trajectories in a unified format. Scalable dynamics learning over such heterogeneous data is enabled by prediction in a structured DINO latent space, which avoids redundant pixel-space appearance modeling. Complementing this representation, LDA-1B employs a multi-modal diffusion transformer to handle asynchronous vision and action streams, enabling stable training at the 1B-parameter scale. Experiments in simulation and the real world show LDA-1B outperforms prior methods (e.g., $\pi_{0.5}$) by up to 21\%, 48\%, and 23\% on contact-rich, dexterous, and long-horizon tasks, respectively. Notably, LDA-1B enables data-efficient fine-tuning, gaining 10\% by leveraging 30\% low-quality trajectories typically harmful and discarded.