Reproduce “PartLLM: A Unified Multimodal Foundation for 3D Part Segmentation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zhe Zhu, Yiheng Zhang, Peng Li, Zixing Zhao, Honghua Chen, Yaqing Zhang, Le Wan, Zhiyang Dou and 4 more

PartLLM unifies text queries, point prompts, and full-shape 3D part segmentation in one prompt-conditioned model. It generates named parts with coarse locations, then jointly assigns shape points to them or background. On 3DCoMPaT200 Fine, it scores 63.1 SA-mIoU, versus 20.8 for PartSAM+MLLM; semantic compatibility is judged by GPT-5.5, so the comparison is not a human evaluation.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.