LLM2R: LLM-Based 3D Part Reassembler via Sequence Modeling
October 12, 2026
LLM2R treats 3D part assembly as a sequence modeling task by using a Point-BERT encoder and Q-Former to convert geometry into tokens for a LoRA-adapted causal LLM. This single-pass approach overcomes the scaling limitations of traditional pairwise or iterative verification methods in complex objects.
HOW THIS AFFECTS YOU
●
builderThis provides a potential alternative to slow iterative assembly pipelines.
●
researcherYou can use sequence modeling paradigms for geometric reconstruction tasks.