EvolvingGrasp: Evolutionary Grasp Generation via Efficient Preference Alignment

ICCV 2025
Yufei Zhu*¹  Yiming Zhong*¹  Zemin Yang¹  Peishan Cong¹  Jingyi Yu¹  Xinge Zhu²  Yuexin Ma^1 
1ShanghaiTech University  ²Chinese University of Hong Kong
*Indicates Equal Contribution

We propose EvolvingGrasp, an evolutionary grasp generation method that continuously enhances grasping performance through efficient preference alignment.

📣 News

• [6/26/2025] EvolvingGrasp has been accepted by ICCV 2025!!!🎉🎉🎉
• [3/19/2025] The paper has been released.

🚩 Plan

• Release Paper.
• Release Inference Code of Consistency Model.
• Release evaluation Code.
• Release training Code.
• Release Checkpoints in different datasets.

😲 Results

Evolution of robotic grasp preferences during efficient feedback-driven finetuning across 10 epochs.

Please refer to our homepage for more thrilling results!

📚 Datasets

Please refer to DexGrasp Anything to process and download each dataset.

🛠️ Setup

• 1. Create a new conda environemnt and activate it.（My CUDA version (nvcc --version) is 11.7）

  conda create -n evolvinggrasp python=3.8
  conda activate evolvinggrasp
  pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 --extra-index-url https://download.pytorch.org/whl/cu113
  pip install peft --no-dependencies

• 2. Install the required packages. You can change TORCH_CUDA_ARCH_LIST according to your GPU architecture.

  TORCH_CUDA_ARCH_LIST="7.0;7.5;8.0;8.6" pip install -r requirements.txt

  Please install in an environment with a GPU, otherwise it will error.

  cd src
  git clone https://github.com/wrc042/CSDF.git
  cd CSDF
  pip install -e .
  cd ..
  git clone https://github.com/facebookresearch/pytorch3d.git
  cd pytorch3d
  git checkout tags/v0.7.2  
  FORCE_CUDA=1  TORCH_CUDA_ARCH_LIST="7.5;8.0;8.6"  python setup.py install
  cd ..

• 3. Install the Isaac Gym Follow the official installation guide to install Isaac Gym and its dependencies. You will get a folder named IsaacGym_Preview_4_Package.tar.gz put it in ./src/IsaacGym_Preview_4_Package.tar.gz

  tar -xzvf IsaacGym_Preview_4_Package.tar.gz
  cd isaacgym/python
  pip install -e .

Before training and testing, please ensure that you set the dataset path, model size, whether to use LLM, sampling method, and other parameters in configs.

📖 Getting Started

Sample and Evaluation

The evaluation is after sampling grasp poses. First, some grasp poses are sampled, then, we will compute quantitative metrics using these sampled results.

cd grasp_gen
bash scripts/grasp_gen_ur/sample_CM.sh ${exp_dir} [OPT]
# e.g., Running without Physics-Guided Sampling:   bash scripts/grasp_gen_ur/sample.sh /outputs/exp_dir [OPT]
# e.g., Running with Physics-Guided Sampling:   bash scripts/grasp_gen_ur/sample.sh /outputs/exp_dir OPT

• [OPT] is an optional parameter for Physics-Guided Sampling.

🎫 License

For academic use, this project is licensed under the 2-clause BSD License.

💓 Acknowledgement

We would like to acknowledge that some codes and datasets are borrowed from Scene-Diffuser, UGG, DexGrasp Anything. We appreciate the authors for their great contributions to the community and for open-sourcing their code and datasets.

🖊️ Citation

If you find our work helpful, please cite:

@article{zhu2025evolvinggrasp,
  title={Evolvinggrasp: Evolutionary grasp generation via efficient preference alignment},
  author={Zhu, Yufei and Zhong, Yiming and Yang, Zemin and Cong, Peishan and Yu, Jingyi and Zhu, Xinge and Ma, Yuexin},
  journal={arXiv preprint arXiv:2503.14329},
  year={2025}
}

Please also consider citing the following papers that inspired EvolvingGrasp.

@inproceedings{motionlcm,
  title={Motionlcm: Real-time controllable motion generation via latent consistency model},
  author={Dai, Wenxun and Chen, Ling-Hao and Wang, Jingbo and Liu, Jinpeng and Dai, Bo and Tang, Yansong},
  booktitle={ECCV},
  pages={390--408},
  year={2025}
}
@article{zhong2025dexgrasp,
  title={DexGrasp Anything: Towards Universal Robotic Dexterous Grasping with Physics Awareness},
  author={Zhong, Yiming and Jiang, Qi and Yu, Jingyi and Ma, Yuexin},
  journal={arXiv preprint arXiv:2503.08257},
  year={2025}
}