Mask2Real-WM: Segmentation Masks as a Sim-to-Real Bridge for Controllable Dexterous World Models

arXiv 2026 42.9 method, system, application

TLDR

Mask2Real-WM uses segmentation masks as a sim-to-real bridge for a controllable dexterous world model, decoupling dynamics and rendering.

Reasoning

The paper presents a novel two-stage world model that leverages segmentation masks to bridge simulation and reality, enabling effective pretraining on synthetic data and fine-tuning on minimal real data. Strengths include the decoupled architecture and demonstrated per-DoF controllability; weaknesses include domain specificity to dexterous manipulation and reliance on mask annotations.

Read-first score

Read-first score 42.9, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 42.

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=baseline,benchmark,evaluation,experiment

Topical relevance 29%

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 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: 351.

Keyword Scores

world model

world dynamics prediction

generative world model

video world model

model-based reinforcement learning world model

interactive world model

world simulator

Deep Analysis

Innovations

Decoupling world model into a dynamics model that predicts future segmentation masks and a rendering model that maps masks to photorealistic RGB
Using segmentation masks as a sim-to-real bridge to enable large-scale pretraining on synthetic simulation data and fine-tuning on real demonstrations
Achieving per-degree-of-freedom action controllability across all 23 DoF via mask conditioning and simulation pretraining

Methodology

Mask2Real-WM is a two-stage action-conditioned world model. The dynamics model predicts future segmentation masks from past masks and 23-DoF action sequences, pretrained on over 50 hours of synthetic data and fine-tuned on less than 2.5 hours of real demonstrations. The rendering model uses a ControlNet-augmented Stable Video Diffusion backbone to generate photorealistic RGB from predicted masks.

Key Results

Mask conditioning and simulation pretraining are both necessary for per-DoF action controllability across all 23 degrees of freedom; monolithic baselines only capture broad hand and end-effector trajectories but not fine-grained per-joint action effects.

Resources

Paper