Deform360: A Massive Multi-view Visuotactile Dataset for Deformable World Models

arXiv 2026 46.5 benchmark

TLDR

A large-scale multi-view visuotactile dataset for evaluating deformable object world models, comparing 2D video and 3D particle approaches.

Reasoning

The paper's strength lies in its massive real-world dataset and systematic comparison of world model paradigms, but it is limited to deformable objects and only provides a preliminary robot planning demonstration rather than a full method.

Read-first score

Read-first score 46.5, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 43.

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=analysis,benchmark,dataset,evaluation,metric

Topical relevance 29%

61.4

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=dataset

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: 576.

Keyword Scores

world model

world dynamics prediction

video world model

world simulator

generative world model

model-based reinforcement learning world model

interactive world model

Deep Analysis

Innovations

Large-scale multi-view visuotactile dataset (Deform360) with 198 objects, 1,980 sequences, 215 hours, 41 cameras, and bimanual tactile grippers
Novel markerless visuotactile 3D tracking pipeline for extracting dense geometry and motion
Systematic comparison of 2D video models and 3D particle models for deformable world modeling
Benchmark and insights into trade-offs between structural priors and scalability

Methodology

Deform360 dataset collected with 41 surround-view cameras and bimanual tactile grippers across 1,980 interaction sequences of 198 daily objects. A markerless visuotactile 3D tracking pipeline extracts dense geometry and motion. State-of-the-art 2D video and 3D particle world models are evaluated on this data, and a robot planning task demonstrates real-world applicability.

Key Results

The evaluation reveals trade-offs between 2D video models and 3D particle models regarding structural priors and scalability. A preliminary robot planning demonstration shows the dataset's potential for real-world deformable object manipulation.

Resources

Paper