Pith. sign in

REVIEW 8 cited by

Efficient Physics Simulation for 3D Scenes via MLLM-Guided Gaussian Splatting

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2411.12789 v3 pith:LXTYXPQE submitted 2024-11-19 cs.CV

Efficient Physics Simulation for 3D Scenes via MLLM-Guided Gaussian Splatting

classification cs.CV
keywords distributionmodelobjectsphysicalanythingcomputationalcostsgeneration
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Recent advancements in 3D generation models have opened new possibilities for simulating dynamic 3D object movements and customizing behaviors, yet creating this content remains challenging. Current methods often require manual assignment of precise physical properties for simulations or rely on video generation models to predict them, which is computationally intensive. In this paper, we rethink the usage of multi-modal large language model (MLLM) in physics-based simulation, and present Sim Anything, a physics-based approach that endows static 3D objects with interactive dynamics. We begin with detailed scene reconstruction and object-level 3D open-vocabulary segmentation, progressing to multi-view image in-painting. Inspired by human visual reasoning, we propose MLLM-based Physical Property Perception (MLLM-P3) to predict mean physical properties of objects in a zero-shot manner. Based on the mean values and the object's geometry, the Material Property Distribution Prediction model (MPDP) model then estimates the full distribution, reformulating the problem as probability distribution estimation to reduce computational costs. Finally, we simulate objects in an open-world scene with particles sampled via the Physical-Geometric Adaptive Sampling (PGAS) strategy, efficiently capturing complex deformations and significantly reducing computational costs. Extensive experiments and user studies demonstrate our Sim Anything achieves more realistic motion than state-of-the-art methods within 2 minutes on a single GPU.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Beyond Thermal Imaging: Inferring Thermophysical Properties from Time-Resolved Thermal Observations

    cs.CV 2026-07 conditional novelty 6.0

    A neural-field plus differentiable FEM pipeline recovers spatially varying thermal diffusivity on reconstructed 3D objects from synthetic thermal sequences and partially transfers to held-out heating/cooling conditions.

  2. DR-GS: Physically-Based Deformable and Relightable 2D Gaussians

    cs.CV 2026-06 unverdicted novelty 6.0

    DR-GS is a unified 2D Gaussian framework that integrates physically-based inverse rendering, relighting, and deformation by explicitly disentangling geometry, illumination, and material representations.

  3. PhyGenHOI: Physically-Aware 4D Generation of Dynamic Human-Object Interactions

    cs.CV 2026-05 unverdicted novelty 6.0

    PhyGenHOI couples a motion diffusion model for humans with material point method simulation for objects on 3D Gaussians, using attraction loss, contact re-simulation, and masked video-SDS to produce physically consist...

  4. MatPhys: Learning Material-Aware Physics Parameters for Deformable Object Simulation from Videos

    cs.CV 2026-05 unverdicted novelty 6.0

    MatPhys is a feed-forward framework that predicts consistent part-level spring-mass parameters for deformable object simulation from monocular videos using semantic decomposition and a material embedding codebook.

  5. CP4D: Compositional Physics-aware 4D Scene Generation

    cs.CV 2026-06 unverdicted novelty 5.0

    CP4D generates physically consistent 4D scenes via compositional integration of pre-trained 3D models, hybrid simulator-diffusion motion synthesis, and automated scene composition.

  6. Learning Action-Conditional and Object-Centric Gaussian Splatting World Models for Rigid Objects

    cs.RO 2026-06 unverdicted novelty 5.0

    MRO-GWM learns action-conditional 3D dynamics of rigid objects via object-centric Gaussians and a transformer, evaluated on synthetic multi-object scenes and in simulation for non-prehensile manipulation.

  7. Occlusion-Robust Multi-Object Decoupling for Physics-Based Robotic Interaction

    cs.CV 2026-06 unverdicted novelty 4.0

    A pipeline combining SAM2 segmentation, 3D Gaussian Splatting, and joint Score Distillation Sampling with 2D/3D diffusion priors reconstructs decoupled multi-object geometries from occluded sparse views for MPM simulation.

  8. Occlusion-Robust Multi-Object Decoupling for Physics-Based Robotic Interaction

    cs.CV 2026-06 unverdicted novelty 4.0

    A new pipeline for occlusion-robust multi-object 3D reconstruction from sparse views supports physics-based robotic interaction.