Xiaomi-Robotics-U0: Unified Embodied Synthesis with World Foundation Model

arXiv 2026 41.6 method, application

TLDR

A 38B multimodal autoregressive model for unified embodied synthesis, achieving SOTA on scene generation and video tasks.

Reasoning

The paper presents a large-scale model that unifies several embodied generation tasks, showing strong empirical results on benchmarks and human evaluations. However, the abstract lacks details on architecture, training data, and ablation studies, making it difficult to assess the novelty and contributions fully.

Read-first score

Read-first score 41.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 32.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,evaluation,metric,result

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=checkpoint,code

Topical relevance 29%

45.7

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Citation impact 18%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 461.

Keyword Scores

world model

generative world model

video world model

world simulator

interactive world model

world dynamics prediction

model-based reinforcement learning world model

Deep Analysis

Innovations

Unified embodied synthesis framework that treats embodied generation as an extension of foundation image and video generation, jointly optimizing text-to-image, editing, embodied scene generation, embodied transfer, and embodied video generation.
First model to support high-quality multi-view scene generation across multiple robot embodiments.
Structured and controllable embodied transfer that enables fine-grained editing while preserving multi-view consistency and interaction dynamics.
Preservation of the generalization of pre-trained world foundation models during adaptation to embodied settings, avoiding the typical sacrifice of visual knowledge.

Methodology

A 38-billion-parameter multimodal autoregressive model that jointly optimizes multiple embodied generation tasks as an extension of foundation image and video generation. The approach preserves the pre-trained world foundation model's generalization while adapting it to robot data, avoiding the common trade-off between embodied performance and large-scale visual knowledge.

Key Results

The model outperforms GPT-Image-2.0 in human evaluations of embodied scene generation and transfer, ranks first on the World Arena benchmark for embodied video generation, and raises the out-of-distribution success rate of pi_0.5 from 36.9% to 63.2% on challenging real-world manipulation tasks.

Resources

Paper