Research map / Detection and segmentation / Open-set and few-shot detection
Differentiable Mathematical Programming for Object-Centric Representation Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 regularized equality-constrained QP min-cut partitioning in place of slot attention | object-centric slot model · Clevr | mIoU: 49.04 vs 36.61 | Helped |
| Use min-cut partitioning in place of slot attention | object-centric slot model · ClevrTex | mIoU and ARI-FG: 42.4 mIoU vs 42.25 GNM; 80.2 ARI-FG vs 79.90 DTI | Helped |
| Use min-cut partitioning in place of slot attention | object-centric slot model · CAMO (eval-only) | mIoU and ARI-FG: 38.9 mIoU vs 27.54 DTI; 75.5 ARI-FG vs 72.90 DTI | Helped |
| Use min-cut partitioning in place of slot attention | object-centric slot model · OOD (eval-only) | mIoU: 38.0 vs 40.84 GNM (best) | Mixed |
| Use min-cut partitioning with pairwise matching in place of slot attention with Hungarian matching | MOVi-A video object discovery · MOVi-A | ARI-FG: 65.6 vs 4.2 | Helped |
| Replace relaxed QP with regularized equality-constrained QP (gamma=0.1) for bipartite matching | Cora bipartite matching · Cora matching validation | average maximization objective: 7.01±1.15 vs 6.15±0.38 | Helped |
| Add a background slot via a separate 2-partition mask multiplied with the object mask (ablation) | min-cut object-centric model · Clevr and ClevrTex | qualitative | No clear effect |
| Use regularized EC-QP instead of exact LP for min-cut (ablation of exact solution not reported) (ablation) | min-cut partitioning · Clevr | ablation | No clear effect |
Closest papers by meaning
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · similarity 0.88)
- Multi-Object Representation Learning via Feature Connectivity and Object-Centric Regularization (NeurIPS 2023 · similarity 0.87)
- Object-Centric Representation Learning for Enhanced 3D Semantic Scene Graph Prediction (NeurIPS 2025 · similarity 0.85)
- Linear Transformer Topological Masking with Graph Random Features (ICLR 2025 · similarity 0.83)
- Continuous Partitioning for Graph-Based Semi-Supervised Learning (NeurIPS 2024 · similarity 0.83)
- Partition First, Embed Later: Laplacian-Based Feature Partitioning for Refined Embedding and Visualization of High-Dimensional Data (ICML 2025 · 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.