PhysMani: Physics-principled 3D World Model for Dynamic Object Manipulation

arXiv 2026 41.2 method, benchmark, application

TLDR

PhysMani couples a physics-principled 3D Gaussian world model with a policy for dynamic object manipulation, achieving superior success in simulation and real-world tasks.

Reasoning

The paper introduces a novel framework integrating physics-principled 3D world modeling with action policy, supported by a new benchmark and real-world experiments. Strengths include clear methodology and empirical validation; weaknesses are not evident from the abstract but limited detail on baselines.

Read-first score

Read-first score 41.2, 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,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: 265.

Keyword Scores

world model

world dynamics prediction

world simulator

model-based reinforcement learning world model

interactive world model

generative world model

video world model

Deep Analysis

Innovations

Physics-principled 3D Gaussian world model with divergence-free Gaussian velocity field for physically grounded future dynamics prediction
Future-aware action policy model that integrates predicted 3D scene dynamics via learnable token-based cross-attention
PhysMani-Bench, a dynamic manipulation benchmark with 16 tasks

Methodology

PhysMani couples a physics-principled 3D Gaussian world model with a future-aware action policy model. The world model learns a divergence-free Gaussian velocity field via online optimization for fast and physically grounded future dynamics prediction. The policy model integrates the predicted 3D scene future dynamics through a learnable token-based cross-attention module.

Key Results

PhysMani achieves superior success rate over strong baselines on the introduced PhysMani-Bench (16 tasks) in both simulation and real-world robot experiments.

Resources

Paper