LLM2R: From Large Language Models to 3D Reassemblers
Overview
LLM2R uses a pretrained decoder-only LLM with custom architecture to assemble 3D parts by predicting SE(3) transformations. It outperforms existing methods on multiple benchmarks, achieving highest accuracy and fastest inference on Breaking Bad. A fracture-specialized method remains more accurate there. LLM2R compresses parts into geometric tokens using Point-BERT and Q-Former, then processes them with LoRA-adapted LLM in one pass. Key innovation: learnable summary tokens for efficient assembly. No deployment or autonomy evidence reported yet. Latest: arXiv:2610.18604v1 (2026-10-14).","latest":"LLM2R achieves highest part accuracy and success rate on all three PartNet categories, best rotation and translation error on PartNeXt, and lowest translation error and fastest inference on Breaking Bad, where only a fracture-specialized method is more accurate."}
Generated from attributed reports · Updated 2 hours ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
No current evidence-backed claims. Missing information remains not reported.
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalLLM2R: From Large Language Models to 3D Reassemblers
LLM2, a novel approach to 3D part assembly, uses a pretrained decoder-only large language model (LLM) with a custom architecture to process parts as a sequence. The method compresses parts into geometric tokens using a Point-BERT encoder and Q-Former, then uses a causal LLM with LoRA adaptation to predict part poses in a single non-autoregressive pass. A key innovation is the use of learnable summary tokens that aggregate the entire assembly, enabling efficient and accurate part assembly. LLM2R achieves the highest part accuracy and success rate on all three PartNet categories, the best rotation and translation error on PartNeXt, and the lowest translation error and fastest inference on Breaking Bad.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).