Research map / Continual learning / Continual fine-tuning of LLMs
Factor-Wise Homogeneity of Slot-Attention for Continual Object-Centric Learning
Ilmin Kang, Hoyong Kim, Seungju Bang, Minwoo Kang, Kangil Kim
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 decoder-only post replay (DPR) | Slot Attention · C-Tetrominoes SST, E0/T1 | FG-ARI: 41.71 -> 98.81 | Improved |
| add DPR | Slot Attention · C-CLEVR SST, E0/T1 | FG-ARI: 63.83 -> 92.00 | Improved |
| add DPR to a model without factor-wise homogeneity (ablation) | SlotMLP · C-Tetrominoes SST, E0/T1 | FG-ARI: 32.85 -> 45.75 | Improved |
| add DPR to a model without factor-wise homogeneity (ablation) | MONet · C-Tetrominoes SST, E0/T1 | FG-ARI: 43.51 -> 44.41 | No clear effect |
| add DPR to a Slot Attention variant with homogeneity (ablation) | BO-QSA · C-Tetrominoes SST, E0/T1 | FG-ARI: 50.46 -> 96.54 | Improved |
| add DPR on novel color factor | Slot Attention · C-Tetrominoes SST-(color), E0/T1 | FG-ARI: 70.45 -> 96.84 | Improved |
| add DPR on shape+color tasks | Slot Attention · C-Tetrominoes SST-(shape+color), E0/T1 | FG-ARI: 48.03 -> 94.19 | Improved |
| add DPR with DINOSAUR on PASCAL VOC | DINOSAUR · PASCAL VOC, |T|=4, E0/T3 | FG-ARI: 14.63 -> 16.69 | Improved |
| add DPR under texture shift | Slot Attention · CLEVR to CLEVR-Tex, E0/T1 | FG-ARI: 68.68 -> 82.75 | Improved |
| add DPR in open-world task | Slot Attention · C-Tetrominoes open-world Tt+1 | FG-ARI: 54.49 -> 98.70 | Improved |
| use slot attention masks instead of decoder masks (ablation) | Slot Attention · C-CLEVR SST, E0/T0 | FG-ARI: 99.86 -> 92.54 | Hurt |
| use slot attention masks instead of decoder masks (ablation) | Slot Attention · C-CLEVR SST, E0/T1 | FG-ARI: 41.71 -> 85.19 | Improved |
Closest papers by meaning
- Adaptive Slot Attention: Object Discovery with Dynamic Slot Number (CVPR 2024 · similarity 0.86)
- Overcoming Catastrophic Forgetting in Visual Continual Learning with Reinforcement Fine-Tuning (NeurIPS 2026 · similarity 0.86)
- Identifiable Object-Centric Representation Learning via Probabilistic Slot Attention (NeurIPS 2024 · similarity 0.86)
- Slot State Space Models (NeurIPS 2024 · similarity 0.85)
- Measuring Representational Shifts in Continual Learning: A Linear Transformation Perspective (ICML 2025 · similarity 0.85)
- Understanding the Forgetting of (Replay-based) Continual Learning via Feature Learning: Angle Matters (ICML 2025 · similarity 0.85)
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.
