AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

BrickNet: Graph-Backed Generative Brick Assembly

Peter Kulits, Cordelia Schmid

CVPR 2026 · Graph learning › Graph self-supervised learning · approach: Diffusion-based graph generation · 5 citations (OpenAlex)

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 changedWhereMeasuredEffect
use graph-backed parametrization instead of pose-basedBrickNet LLM (14b) · BrickNet unconditional, nucleus samplingsteps until connectivity failure: 49.9 -> 96.9Improved
use graph-backed parametrization instead of pose-basedBrickNet LLM (14b) · BrickNet unconditional, nucleus samplingsteps until collision: 19.1 vs 22.4No clear effect
switch ancestral to nucleus samplingGraph-backed model (0.6b) · BrickNet unconditionalsteps until connectivity failure: 45.0 -> 94.1Improved
scale model size from 0.6b to 14bGraph-backed model · BrickNet unconditional, nucleus samplingsteps until connectivity failure: 94.1 -> 96.9Improved
scale model size from 1.7b to 14bBrickNet SFT · BrickNet text-conditionedVQAScore: 0.59318309911395772 vs 0.60203790253581246Mixed
add unconditional pretraining before SFT (14b) (ablation)BrickNet SFT · BrickNet held-out perplexityperplexity: 1.298 -> 1.266Improved
add unconditional pretraining before SFT (0.6b) (ablation)BrickNet SFT · BrickNet held-out perplexityperplexity: 1.343 -> 1.298Improved
add more SFT data (None to Full, with full PT) (ablation)BrickNet 14b · BrickNet held-out perplexityperplexity: 1.300 -> 1.266Improved
add more PT data (None to Full SFT, quarter PT vs full PT) (ablation)BrickNet 14b · BrickNet held-out perplexityperplexity: 1.361 -> 1.300Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy