Research map / Detection and segmentation / Open-set and few-shot detection
QASA: Quality-Guided K-Adaptive Slot Attention for Unsupervised Object-Centric Learning
Tianran Ouyang, Xingping Dong, Jing Zhang, Mang Ye, Kaihao Zhang, Bo Du
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 |
|---|---|---|---|
| add coverage criterion to slot selection (ablation) | QASA (Transformer) · COCO | mBOi: 25.5 -> 25.3 | No clear effect |
| add coverage criterion to slot selection (ablation) | QASA (Transformer) · COCO | mBOc: 25.9 -> 34.5 | Improved |
| add slot-quality prioritization on top of coverage (ablation) | QASA (Transformer) · COCO | mBOi: 25.3 -> 35.0 | Improved |
| add novelty constraint on top of coverage and quality (ablation) | QASA (Transformer) · COCO | mBOi: 35.0 -> 36.7 | Improved |
| add K/V gate g1 to decoder (ablation) | QASA (Transformer) · COCO | mBOi: 25.5 -> 33.2 | Improved |
| add logit gate g2 alone (ablation) | QASA (Transformer) · COCO | mBOi: 25.5 -> 25.5 | No clear effect |
| add logit gate g2 on top of g1 (ablation) | QASA (Transformer) · COCO | mBOi: 33.2 -> 36.7 | Improved |
| use Transformer instead of MLP decoder | QASA · COCO | mBOi: 33.6 -> 36.7 | Improved |
| increase Kmax from 20 to 32 beyond max object count (ablation) | QASA (Transformer) · MOVi-C | mBOi: 46.9 -> 45.4 | Hurt |
| reduce Kmax from 20 to 15 (ablation) | QASA (Transformer) · MOVi-C | mBOi: 46.9 -> 44.4 | Hurt |
| increase Kmax from 30 to 64 (ablation) | QASA (Transformer) · MOVi-E | mBOi: 39.1 -> 38.7 | Hurt |
| use QASA slots instead of SPOT for property prediction (MLP predictor) | QASA (Transformer) · COCO property prediction | class top-1: 45.9 -> 59.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MetaSlot — MetaSlot: Break Through the Fixed Number of Slots in Object-Centric Learning (NeurIPS 2025) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Adaptive Slot Attention: Object Discovery with Dynamic Slot Number (CVPR 2024 · similarity 0.91)
- Grounded Object-Centric Learning (ICLR 2024 · similarity 0.88)
- Slot-Guided Adaptation of Pre-trained Diffusion Models for Object-Centric Learning and Compositional Generation (ICLR 2025 · similarity 0.88)
- MetaSlot: Break Through the Fixed Number of Slots in Object-Centric Learning (NeurIPS 2025 · similarity 0.88)
- Identifiable Object-Centric Representation Learning via Probabilistic Slot Attention (NeurIPS 2024 · similarity 0.87)
- GLASS: Guided Latent Slot Diffusion for Object-Centric Learning (CVPR 2025 · similarity 0.87)
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.
