Salvato in:
| Autori principali: | , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2602.12215 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918334731124736 |
|---|---|
| author | Lyu, Jiangran Liu, Kai Zhang, Xuheng Liao, Haoran Feng, Yusen Zhu, Wenxuan Shen, Tingrui Chen, Jiayi Zhang, Jiazhao Dong, Yifei Cui, Wenbo Qi, Senmao Wang, Shuo Zheng, Yixin Yan, Mi Shi, Xuesong Li, Haoran Zhao, Dongbin Liu, Ming-Yu Zhang, Zhizheng Yi, Li Wang, Yizhou Wang, He |
| author_facet | Lyu, Jiangran Liu, Kai Zhang, Xuheng Liao, Haoran Feng, Yusen Zhu, Wenxuan Shen, Tingrui Chen, Jiayi Zhang, Jiazhao Dong, Yifei Cui, Wenbo Qi, Senmao Wang, Shuo Zheng, Yixin Yan, Mi Shi, Xuesong Li, Haoran Zhao, Dongbin Liu, Ming-Yu Zhang, Zhizheng Yi, Li Wang, Yizhou Wang, He |
| contents | 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., $π_{0.5}$) 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_12215 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LDA-1B: Scaling Latent Dynamics Action Model via Universal Embodied Data Ingestion Lyu, Jiangran Liu, Kai Zhang, Xuheng Liao, Haoran Feng, Yusen Zhu, Wenxuan Shen, Tingrui Chen, Jiayi Zhang, Jiazhao Dong, Yifei Cui, Wenbo Qi, Senmao Wang, Shuo Zheng, Yixin Yan, Mi Shi, Xuesong Li, Haoran Zhao, Dongbin Liu, Ming-Yu Zhang, Zhizheng Yi, Li Wang, Yizhou Wang, He Robotics 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., $π_{0.5}$) 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. |
| title | LDA-1B: Scaling Latent Dynamics Action Model via Universal Embodied Data Ingestion |
| topic | Robotics |
| url | https://arxiv.org/abs/2602.12215 |