SANA-WM: Efficient Minute-Scale World Modeling with Hybrid Linear Diffusion Transformer

arXiv 2026 55.8 method, system

TLDR

SANA-WM is an efficient 2.6B-parameter open-source world model for minute-scale 720p video generation with camera control, achieving high quality and efficiency.

Reasoning

The paper presents a novel architecture with hybrid linear attention and dual-branch camera control, demonstrating strong efficiency in data, training, and inference. However, it is limited to video generation and does not address interactive or RL-based world modeling, and evaluation is only on a custom benchmark.

Read-first score

Read-first score 55.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 44.

Recency 6%

100

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

Citation impact 18%

77.9

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.77929302

Methodology quality 18%

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

Topical relevance 29%

62.9

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

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Citation velocity 12%

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

Field roles

FoundationFrontierBridgeMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 323.

Keyword Scores

world model

video world model

generative world model

world dynamics prediction

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Hybrid Linear Attention combining frame-wise Gated DeltaNet (GDN) with softmax attention for memory-efficient long-context modeling
Dual-Branch Camera Control ensuring precise 6-DoF trajectory adherence
Two-Stage Generation Pipeline applying a long-video refiner to stage-1 outputs for improved quality and consistency
Robust Annotation Pipeline extracting accurate metric-scale 6-DoF camera poses from public videos to yield high-quality action labels

Methodology

SANA-WM is a 2.6B-parameter hybrid linear diffusion transformer trained natively for one-minute video generation. It uses approximately 213K public video clips with metric-scale pose supervision, completes training in 15 days on 64 H100 GPUs, and is evaluated on a one-minute world-model benchmark against baselines such as LingBot-World and HY-WorldPlay. The model incorporates a distilled variant with NVFP4 quantization for deployment on a single RTX 5090.

Key Results

SANA-WM achieves visual quality comparable to large-scale industrial baselines (LingBot-World, HY-WorldPlay) while offering 36× higher throughput, and demonstrates stronger action-following accuracy than prior open-source baselines. Its distilled variant can denoise a 60-second 720p clip on a single RTX 5090 in 34 seconds.

Resources

Paper