Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
A causal view of compositional zero-shot recognition
Yuval Atzmon, Felix Kreuk, Uri Shalit, Gal Chechik
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 |
|---|---|---|---|
| model stable attribute and object core features and infer pairs by the conditional image likelihood | Causal · AO-CLEVr 5:5 split | Unseen accuracy: 26% to 47% | Improved |
| model stable attribute and object core features and infer pairs by the conditional image likelihood | Causal · AO-CLEVr 5:5 split | Seen accuracy: 86% to 84% | Hurt |
| add conditional-independence regularization to VisProd | VisProd · AO-CLEVr 5:5 split | Unseen accuracy: 19% to 38% | Improved |
| add conditional-independence regularization to VisProd | VisProd · AO-CLEVr 5:5 split | Seen accuracy: 85% to 82% | Hurt |
| add the independence loss (ablation) | Causal · Zappos test | Unseen accuracy: 22.5 ± 2.0 -> 26.6 ± 1.6 | Improved |
| add the independence loss (ablation) | Causal · Zappos test | Seen accuracy: 45.5 ± 3.7 -> 39.7 ± 2.2 | Hurt |
| add the independence loss (ablation) | Causal · Zappos test | Harmonic: 29.4 ± 1.5 -> 31.8 ± 1.7 | Improved |
| use non-alternating rather than alternating optimization (ablation) | AO-CLEVr Causal · AO-CLEVr | overall results | Mixed |
| remove frequency weighting from the attribute and object reconstruction losses (ablation) | AO-CLEVr Causal · AO-CLEVr | Harmonic: 68.8% to 67.6% | Hurt |
| remove frequency weighting from the attribute and object reconstruction losses (ablation) | AO-CLEVr Causal · AO-CLEVr | Unseen accuracy: 57.5% to 56.2% | Hurt |
Closest papers by meaning
- Does Data Scaling Lead to Visual Compositional Generalization? (ICML 2025 · similarity 0.89)
- Revealing the Proximate Long-Tail Distribution in Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.88)
- A Dynamic Learning Method towards Realistic Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.87)
- Retrieval-Augmented Primitive Representations for Compositional Zero-Shot Learning (AAAI 2024 · similarity 0.87)
- Beyond Seen Primitive Concepts and Attribute-Object Compositional Learning (CVPR 2024 · similarity 0.87)
- Scalable Evaluation and Neural Models for Compositional Generalization (NeurIPS 2025 · similarity 0.86)
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.
