Rollback-Free Stable Brick Structures Generation
For 3D generation of brick structures, this method eliminates the need for external simulators during inference, dramatically improving efficiency.
This work tackles the problem of generating physically stable brick structures, which previously required slow test-time rollbacks. The proposed reinforcement learning paradigm achieves rollback-free generation, accelerating inference by orders of magnitude while achieving state-of-the-art quality.
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.