Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..

PhysX-3D: Physical-Grounded 3D Asset Generation

Authors: Ziang Cao, Zhaoxi Chen, Liang Pan, Ziwei Liu

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments validate the superior performance and promising generalization capability of our framework. All the code, data, and models will be released to facilitate future research in generative physical AI.
Researcher Affiliation Collaboration Ziang Cao1 Zhaoxi Chen1 Liang Pan2 Ziwei Liu1 1Nanyang Technological University 2Shanghai AI Lab https://physx-3d.github.io/ Corresponding author, EMAIL
Pseudocode No The paper describes methods through textual descriptions and diagrams but does not contain a clearly labeled pseudocode or algorithm block.
Open Source Code Yes All the code, data, and models will be released to facilitate future research in generative physical AI. (...) Answer: [Yes] Justification: We release the code of our generative model and examples of our dataset in the supplementary. We will release the full data as soon as possible.
Open Datasets Yes 1) To bridge the critical gap in physics-annotated 3D datasets, we present Phys XNet the first physics-grounded 3D dataset systematically annotated across five foundational dimensions: absolute scale, material, affordance, kinematics, and function description. (...) Answer: [Yes] Justification: We release the code of our generative model and examples of our dataset in the supplementary. We will release the full data as soon as possible.
Dataset Splits Yes In our experiments, we partition Phys XNet dataset into 24K training samples, 1K validation samples, and 1K test cases.
Hardware Specification Yes Our Phys XGen is trained on 8 NVIDIA A100 GPUS.
Software Dependencies No The paper mentions software components like 'GPT-4o', 'CLIP model [32]', and 'DINOv2' but does not provide specific version numbers for general software dependencies like Python, PyTorch, or CUDA.
Experiment Setup Yes During the VAE and diffusion model training, we adopt Adam W with an initial learning rate of 1 10 4 to optimize the models.