Research map / Label-efficient and robust learning / Recognition and long-tail
One-Bit Active Query With Contrastive Pairs
Yuhang Zhang, Xiaopeng Zhang, Lingxi Xie, Jie Li, Robert C. Qiu, Hengtong Hu, Qi Tian
Check 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 ObCP instead of VAAL at 10% labels | VGG16 active learning on ImageNet · ImageNet | accuracy: 61.3 vs 53.3 | Improved |
| Use ObCP instead of conventional one-bit supervision | CIFAR-100 one-bit active learning · CIFAR-100 | accuracy: 59.5% to 62.1% | Improved |
| Use ObCP instead of conventional one-bit supervision | ImageNet one-bit active learning · ImageNet | accuracy: 60.4% to 63.8% | Improved |
| Add negative contrastive learning for incorrect predictions (ablation) | ObCP · ImageNet-100 | top-1 accuracy: +0.8% | Improved |
| Add the negative loss for incorrect predictions (ablation) | ObCP · ImageNet-100 | top-1 accuracy: +2.2% | Improved |
| Add correct queried labels as positives in the contrastive loss (ablation) | ObCP · ImageNet-100 | top-1 accuracy: +2.5% | Improved |
| Add correct queried labels to the supervised loss (ablation) | ObCP · ImageNet-100 | top-1 accuracy: +6.4% | Improved |
| Spread the query budget across four rounds instead of using it at once (ablation) | ObCP · ImageNet-100 | accuracy: ∼6% | Improved |
| Increase query rounds beyond the selected four (ablation) | ObCP · ImageNet-100 | accuracy | No clear effect |
| Increase the initial fully labeled set size while holding the total budget fixed (ablation) | ObCP · ImageNet-100 | accuracy | No clear effect |
| Use the proposed training-prediction uncertainty metric instead of Random selection (ablation) | ObCP sample selection · CIFAR-100 (30% label budget) | accuracy: ∼4.1% | Improved |
| Use the proposed training-prediction uncertainty metric instead of max-confidence selection (ablation) | ObCP sample selection · CIFAR-100 (30% label budget) | accuracy: ∼2.3% | Improved |
Closest papers by meaning
- Contrastive Active Learning Under Class Distribution Mismatch (TPAMI 2022 · similarity 0.86)
- One-shot Active Learning Based on Lewis Weight Sampling for Multiple Deep Models (ICLR 2024 · similarity 0.84)
- On the Similarities of Embeddings in Contrastive Learning (ICML 2025 · similarity 0.84)
- Active Classification with Few Queries under Misspecification (NeurIPS 2024 · similarity 0.84)
- Active Learning From Positive and Unlabeled Examples (NeurIPS 2026 · similarity 0.84)
- Uncertainty-Aware Contrastive Learning with Hard Negative Sampling for Code Search Tasks (AAAI 2025 · similarity 0.83)
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.
