MegaParts: Scaling Part-Aware 3D Object Generation to 300 Parts via Token-Efficient Autoregressive Modeling

2026-08-14

AI Summary

This study addresses the substantial token overhead and memory bottlenecks inherent in part-aware 3D generation by proposing MegaParts, a novel framework featuring an adaptive-length part tokenizer and unified structured sequence modeling. By integrating vector quantization with efficient long-context training strategies, this approach overcomes existing scalability limitations. The method successfully supports sequences of up to 256k tokens and the generation of objects comprising 300 parts. Experimental results demonstrate that MegaParts achieves superior mesh quality compared to baselines while effectively balancing scalability and high fidelity in large-scale 3D generation. Consequently, this work resolves critical extensibility challenges associated with synthesizing complex objects, marking a significant advancement in scalable part-aware generative modeling.

Abstract

Part-aware 3D object generation is essential for graphics applications such as controllable modeling, editing, and articulation, where objects are represented as coherent assemblies of semantic parts. However, existing part-aware generation methods, do not scale well to highly complex objects. As the number of parts increases, generating detailed geometry becomes prohibitively expensive in token length and memory. We introduce MegaParts, a scalable autoregressive 3D generation framework to address this challenge by combining structured sequence modeling with a token-efficient vector-quantized shape tokenizer. Our tokenizer learns discrete latent representations for part-level geometry by minimizing token usage subject to high-fidelity reconstruction, enabling adaptive-length tokenization based on geometric complexity. On top of this compact representation, we train a large language model to generate object bounding boxes, part bounding boxes, and part shape tokens within a unified structured sequence. Combined with efficient long-context training strategy, our token-efficient formulation scales to objects with up to 300 parts and sequence lengths up to 256k tokens. This substantially extends the scale of part-aware 3D generation while preserving compositional structure and enabling fine-grained part-level control. Our method achieves higher mesh quality than baseline autoregressive and diffusion models, showing that compressed discrete part tokens improve not only scalability but also the achievable fidelity of generated geometry. These results suggest that LLM native token-efficient autoregressive modeling is a compelling alternative to diffusion for large-scale part-aware 3D generation. The project page is available at https://expmaster.github.io/megaparts_webpage.

Problem

• Part-aware 3D object generation
• Scalability
• Token efficiency
• Complex objects
• Memory cost

Innovation

• Token-Efficient Autoregressive Modeling
• Part-Aware 3D Generation
• Vector-Quantized Shape Tokenizer
• Structured Sequence Modeling
• Adaptive-Length Tokenization