Brick-Composer: Using MLLMs for Assembly with Diverse Bricks

In three linesBrick-Composer trains MLLMs to assemble objects from reusable bricks. Authors introduce BC-Bench, an evaluation benchmark, and propose a method combining human design demonstrations, visual/physical feedback, and synthetic experience. Qwen-3-8B achieves 42% step-level success after training, up from <1% baseline.

Summary generated by Claude — human-verified