ResearchOct 21, 2026, 00:00 UTC

Paper introduces LLM2R, using a pretrained LLM to solve 3D part assembly

LLM2R treats part assembly as sequence modeling with a causal LLM and learnable summary tokens, predicting all part poses in one non-autoregressive pass. It reports top results on PartNet, PartNeXt and Breaking Bad, with gains growing as part count rises.

ModelLLM2R
Topicresearch
Topicrobotics

1outlet ran itRead at Hugging Face Papers

Status: Research
Published: Oct 21, 2026, 00:00 UTC That day
Outlets: 1
Written by: Claude How

Every outlet that ran it 1 article

Oct 21, 2026, 00:00 UTC
Hugging Face PapersResearchLead
LLM2R: From Large Language Models to 3D Reassemblers

All news