Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Zero-Shot Learning - the Good, the Bad and the Ugly
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 |
|---|---|---|---|
| Exclude zero-shot test classes that occur in ImageNet 1K pretraining and evaluate on the proposed split | Zero-shot methods using ResNet features · AWA | top-1 accuracy: 78.6 -> 59.9 | Hurt |
| Exclude zero-shot test classes that occur in ImageNet 1K pretraining and evaluate on the proposed split | Zero-shot methods using ResNet features · CUB | top-1 accuracy: 55.3 -> 53.9 | No clear effect |
| Exclude zero-shot test classes that occur in ImageNet 1K pretraining and evaluate on the proposed split | Zero-shot methods using ResNet features · SUN | top-1 accuracy: 59.1 -> 58.1 | No clear effect |
| Add novelty detection (ablation) | CMT · SUN, generalized zero-shot PS | harmonic mean: 11.8 -> 13.3 | Helped |
| Add novelty detection (ablation) | CMT · CUB, generalized zero-shot PS | harmonic mean: 12.6 -> 8.7 | Hurt |
| Add novelty detection (ablation) | CMT · AWA, generalized zero-shot PS | harmonic mean: 1.8 -> 15.3 | Helped |
| Add novelty detection (ablation) | CMT · aPY, generalized zero-shot PS | harmonic mean: 2.8 -> 19.0 | Helped |
| Include seen training classes in the test search space | Zero-shot methods · Generalized zero-shot PS | top-1 accuracy on unseen classes | Hurt |
| Use max-margin bilinear compatibility learning rather than independent class/attribute classifiers or hybrid models | Zero-shot methods · Proposed split (PS) | top-1 accuracy | Helped |
Closest papers by meaning
- A Generalization Theory for Zero-Shot Prediction (ICML 2025 · similarity 0.87)
- Zero-Shot Robustification of Zero-Shot Models (ICLR 2024 · similarity 0.87)
- No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance (NeurIPS 2024 · similarity 0.87)
- Class Distribution Shifts in Zero-Shot Learning: Learning Robust Representations (NeurIPS 2024 · similarity 0.86)
- Improving Generalized Zero-Shot Learning by Exploring the Diverse Semantics from External Class Names (CVPR 2024 · similarity 0.86)
- Zero-Shot Reinforcement Learning from Low Quality Data (NeurIPS 2024 · 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.