Research map / Large language models / AI and society
Toward Few-Shot Learning in the Open World: A Review and Beyond
Hui Xue, Yuexuan An, Yongchun Qin, Wenqian Li, Yixin Wu, Yongjuan Che, Pengfei Fang, Min-Ling Zhang
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 |
|---|---|---|---|
| use metric-based weighting (PCL) instead of gradient-based Nested-MAML | noisy few-shot learning · miniImageNet 5-shot, 40% noise | accuracy: 56.46 -> 67.64 | Improved |
| raise symmetric support-set noise from 0% to 40% | DETA · miniImageNet 5-shot | accuracy: 70.21 -> 57.04 | Hurt |
| raise symmetric support-set noise from 0% to 40% | PCL · miniImageNet 5-shot | accuracy: 70.31 -> 67.64 | Hurt |
| raise symmetric support-set noise from 0% to 40% | RNNP · miniImageNet 5-shot | accuracy: 68.38 -> 51.62 | Hurt |
| use test-time adaptation (DETA) instead of meta-training weighting (IDEAL) | sample-selection NFSL · miniImageNet 5-shot, 40% noise | accuracy: 56.45 -> 57.04 | Mixed |
| add class semantics to model the unknown class (ATT-G vs ATT) | ATT · miniImageNet 1-shot open-set | AUROC: 71.35 -> 72.41 | Improved |
| add global energy open scores (GEL) | few-shot open-set recognition · miniImageNet 5-shot | AUROC: 82.29 vs 81.04 | Improved |
| use the smoothed variant ProCAMsm instead of ProCAM | ProCAM · miniImageNet 1-shot open-set | AUROC: 71.41 -> 71.09 | No clear effect |
| use CLIP language-image pretraining (CPE-CLIP) | incremental few-shot learning · CIFAR-100 incremental | performance drop (PD): 21.05 -> 7.31 | Improved |
| add unlabeled samples (UadCE) to data-based incremental learning | data-strategy IFSL · CUB-200 incremental | performance drop (PD): 14.45 vs 22.04 | Improved |
| use attention fusion across base and novel classes (BiDist) | network-ensemble IFSL · CIFAR-100 incremental | performance drop (PD): 22.89 vs 34.73 | Improved |
Closest papers by meaning
- Scaling Few-Shot Learning for the Open World (AAAI 2024 · similarity 0.91)
- Bayesian Embeddings for Few-Shot Open World Recognition (TPAMI 2022 · similarity 0.90)
- Bongard-OpenWorld: Few-Shot Reasoning for Free-form Visual Concepts in the Real World (ICLR 2024 · similarity 0.86)
- The Surprising Effectiveness of Test-Time Training for Few-Shot Learning (ICML 2025 · similarity 0.86)
- Many-Shot In-Context Learning (NeurIPS 2024 · similarity 0.86)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · 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.
