Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Break and Make: Interactive Structural Understanding Using LEGO Bricks
Aaron Walsman, Muru Zhang, Klemen Kotar, Karthik Desingh, Ali Farhadi, Dieter Fox
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 StudNet-B transformer instead of LSTM | Break and Make agent · Random Construction (8 bricks) | F1a: 0.02 -> 0.12 | Improved |
| use StudNet-A transformer instead of LSTM | Break and Make agent · Random Construction (2 bricks) | F1a: 0.43 -> 0.58 | Improved |
| use StudNet-A instead of LSTM | Break and Make agent · Sliced OMR (2 bricks) | F1a: 0.31 -> 0.00 | Hurt |
| replace StudNet-A decoder with convolutional StudNet-B decoder | StudNet · Sliced OMR (2 bricks) | F1a: 0.00 -> 0.29 | Improved |
| pretrain the ResNet-18 backbone on dense brick shape and color labeling | LSTM · Random Construction (4 bricks) | F1b: 0.41 -> 0.25 | Hurt |
| pretrain the ResNet-18 backbone on dense brick shape and color labeling | LSTM · Random Construction (2 bricks) | F1b: 0.61 -> 0.70 | Improved |
| increase scene size from 2 to 8 bricks | LSTM · Random Construction | F1a: 0.43 -> 0.02 | Hurt |
| increase scene size from 2 to 8 bricks | StudNet-B · Random Construction | F1a: 0.57 -> 0.12 | Hurt |
| switch from random construction to real OMR models | StudNet-B · 2 bricks | F1a: 0.57 -> 0.29 | Hurt |
| switch from random construction to real OMR models | LSTM · 2 bricks | F1a: 0.43 -> 0.31 | Hurt |
Closest papers by meaning
- Learning to Build by Building Your Own Instructions (ECCV 2024 · similarity 0.89)
- BrickNet: Graph-Backed Generative Brick Assembly (CVPR 2026 · similarity 0.88)
- BrickAnything: Geometry-Conditioned Buildable Brick Generation with Structure-Aware Tokenization (NeurIPS 2026 · similarity 0.85)
- LEIA: Learned Environment for Interactive Architected Materials (NeurIPS 2026 · similarity 0.83)
- PhyBlock: A Progressive Benchmark for Physical Understanding and Planning via 3D Block Assembly (NeurIPS 2025 · similarity 0.82)
- IKEA Manuals at Work: 4D Grounding of Assembly Instructions on Internet Videos (NeurIPS 2024 · similarity 0.82)
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.
