用强化学习生成不需回滚的稳定积木结构，效率提升显著
Rollback-Free Stable Brick Structures Generation
- 训练时优化物理约束，无需推理时反复试错
- 生成结构稳定性达当前最佳，推理速度提升数量级
- 适合需要高效生成物理合理3D结构的场景
自回归模型虽已推动3D生成发展，但构建符合重力与连接性要求的稳定积木结构仍具挑战。现有方法依赖外部物理模拟器在推理时进行拒绝采样和逐块回滚，严重限制效率。为此，我们提出一种强化学习范式，将物理有效性验证从测试时修正转移到训练时策略优化。通过使用组装级奖励，模型优化碰撞规避、全局连通性、结构咬合与形状一致性。该范式使模型内化物理先验，首次实现无回滚的稳定积木结构生成。实验表明，本方法在生成质量上达到当前最优，同时推理速度提升数量级。代码与数据集见https://github.com/miniHuiHui/STABLE，模型见https://huggingface.co/miniHui/STABLE。
原文摘要 · Abstract (English)
While autoregressive models have advanced 3D generation, creating physically stable brick structures remains a challenge due to the strict requirements of gravity and interconnectivity. Existing approaches rely on external physical simulators during inference to perform rejection sampling and brick-by-brick rollbacks, which severely bottlenecks efficiency. To address this, we propose a reinforcement learning paradigm that shifts physical validity enforcement from test-time correction to training-time policy optimization. By utilizing assembly-level rewards, the model optimizes for collision avoidance, global connectivity, structural interlocking, and shape conformity. This paradigm allows the model to internalize physical priors, enabling the first rollback-free generation of stable brick structures. Experimental results demonstrate that our approach achieves state-of-the-art generation quality while accelerating inference speed by orders of magnitude. Our code and dataset are available at https://github.com/miniHuiHui/STABLE. Our models are available at https://huggingface.co/miniHui/STABLE.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。