AI Paper Atlas

Research map / 3D vision / 3D generation

3D-aware Disentangled Representation for Compositional Reinforcement Learning

Sungbin Mun, Younghwan Lee, Cheol-Hui Min, Mineui Hong, Young Min Kim

ICLR 2026 · 3D vision › 3D generation · approach: Semantic scene completion

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 3D block-slot attention with factor binding instead of permutation-invariant OSRT slotsOSRT · Clevr3DFG-ARI: 0.365 -> 0.942Improved
use 3D block-slot attention instead of OSRT slotsOSRT · IsaacGym3DFG-ARI: 0.321 -> 0.619Improved
use 3D block-slot attention instead of OSRT slotsOSRT · IsaacGym3DPSNR: 27.35 -> 26.55No clear effect
use 3D block-slot attention instead of OSRT slotsOSRT · Clevr3DDCI informativeness: 0.452 -> 0.844Improved
replace entity-wise EIT policy with block transformer on block-slot representation (ablation)Ours w/ EIT · CG (same color)success rate: 0.682 -> 0.837Improved
replace entity-wise EIT policy with block transformer (ablation)Ours w/ EIT · CGsuccess rate: 0.773 -> 0.895Improved
replace entity-wise EIT policy with block transformer (ablation)Ours w/ EIT · IDsuccess rate: 0.984 -> 0.967No clear effect
replace 2D DLPv2 representation with 3D block-slot representation (EIT policy)DLPv2 w/ EIT · CG (same color)success rate: 0.388 -> 0.682Improved
evaluate on unseen viewpoints, 2D baseline vs 3D block-slot with BTDLPv2 w/ EIT · OOD multi-view ID goalssuccess rate: 0.059 -> 0.948Improved
use single view at inference instead of three views (ablation)3D block-slot representation · IsaacGym3DPSNR: 26.55 -> 25.90Hurt
use object-centric representation instead of VAE with MLP policyVAE w/ MLP · IsaacGym pushsuccess rate: 0.042 -> 0.967Improved

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