Research map / Reinforcement learning / Visual and representation RL
Brick-by-Brick: Combinatorial Construction with Deep Reinforcement Learning
Hyunsoo Chung, Jungtaek Kim, Boris Knyazev, Jinhwi Lee, Graham W. Taylor, Jaesik Park, Minsu Cho
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use the pretrained validity prediction network instead of the jointly trained network (ablation) | Brick-by-Brick action validity prediction · Action validity prediction test set, pivot | precision: 0.9809 -> 0.9909 | Improved |
| Use the pretrained validity prediction network instead of the jointly trained network (ablation) | Brick-by-Brick action validity prediction · Action validity prediction test set, offset | precision: 0.8674 -> 0.9125 | Improved |
| Add node features to the validity prediction GNN (ablation) | Brick-by-Brick action validity prediction · Action validity prediction test set, offset | precision: 0.7931 -> 0.9125 | Improved |
| Add edge features to the validity prediction GNN (ablation) | Brick-by-Brick action validity prediction · Action validity prediction test set, offset | precision: 0.8897 -> 0.9125 | Improved |
| Use a GNN instead of an MLP for action validity prediction | Brick-by-Brick action validity prediction · Action validity prediction test set, offset | precision: 0.5130 -> 0.9125 | Improved |
| Use graph message passing rather than independent MLP brick features (ablation) | Brick-by-Brick construction · Randomly-assembled object construction | episode return | No clear effect |
| Use reinforcement learning rather than sequence-level supervised learning | Brick-by-Brick construction · Randomly-assembled object construction test set | generalization to unseen images | Improved |
| Use image-conditioned Brick-by-Brick instead of Bayesian optimization | Brick-by-Brick construction · ModelNet construction | episode return | Improved |
Closest papers by meaning
- BrickNet: Graph-Backed Generative Brick Assembly (CVPR 2026 · similarity 0.88)
- BricksRL: A Platform for Democratizing Robotics and Reinforcement Learning Research and Education with LEGO (NeurIPS 2024 · similarity 0.85)
- BrickAnything: Geometry-Conditioned Buildable Brick Generation with Structure-Aware Tokenization (NeurIPS 2026 · similarity 0.84)
- Learning Stackable and Skippable LEGO Bricks for Efficient, Reconfigurable, and Variable-Resolution Diffusion Modeling (ICLR 2024 · similarity 0.84)
- Structured Reinforcement Learning for Combinatorial Decision-Making (NeurIPS 2025 · similarity 0.84)
- Constrained Goal-directed Planar Graph Generation with Grammar-based Reinforcement Learning (NeurIPS 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
