Research map / Vision-language and multimodal / CLIP and zero-shot
Open World Compositional Zero-Shot Learning
Check 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 |
|---|---|---|---|
| Add feasibility-based margins to cosine logits (ablation) | CompCos · MIT states validation | AUC: 1.7 to 2.0 | Helped |
| Linearly warm up the feasibility margin α (ablation) | CompCos with feasibility margins · MIT states validation | AUC: 2.0 to 2.1 | Helped |
| Estimate feasibility from state and object similarities rather than either primitive alone (ablation) | CompCos · MIT states validation | AUC: 1.9 to 2.1 | Helped |
| Combine state and object feasibility using an average rather than maximum (ablation) | CompCos · MIT states validation | AUC: 2.0 to 2.1 | Helped |
| Apply feasibility-based masking at inference (ablation) | Closed-world CompCos · MIT states validation | AUC: 1.2 to 1.6 | Helped |
| Apply feasibility-based masking at inference (ablation) | CompCos · MIT states validation | AUC: 2.1 to 2.2 | Helped |
| Apply feasibility-based masking at inference (ablation) | LE+ · MIT states validation | AUC: 0.3 to 0.5 | Helped |
| Apply feasibility-based masking at inference (ablation) | TMN · MIT states validation | AUC: 0.1 to 0.4 | Helped |
| Apply feasibility-based masking at inference (ablation) | SymNet · MIT states validation | AUC: 1.2 to 1.2 | No clear effect |
| Keep the lower-scoring half of unseen compositions instead of the higher-scoring half (ablation) | CompCos · MIT states validation | AUC: 2.2 to 0.06 | Hurt |
| Use CompCos feasibility-based margins and masking instead of SymNet | Compositional zero-shot learning · MIT states open world test | AUC: 0.8 to 1.6 | Helped |
| Use CompCos instead of SymNet | Compositional zero-shot learning · UT Zappos open world test | AUC: 18.5 to 21.3 | Helped |
Closest papers by meaning
- Beyond Seen Primitive Concepts and Attribute-Object Compositional Learning (CVPR 2024 · similarity 0.90)
- Retrieval-Augmented Primitive Representations for Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.90)
- A Dynamic Learning Method towards Realistic Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.90)
- Context-based and Diversity-driven Specificity in Compositional Zero-Shot Learning (CVPR 2024 · similarity 0.90)
- Troika: Multi-Path Cross-Modal Traction for Compositional Zero-Shot Learning (CVPR 2024 · similarity 0.90)
- Revealing the Proximate Long-Tail Distribution in Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.89)
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.