Research map / Label-efficient and robust learning / Recognition and long-tail
Bidirectional Uncertainty-Based Active Learning for Open-Set Annotation
Chen-Chen Zong, Ye-Wen Wang, Kun-Peng Ning, Hai-Bo Ye, Sheng-Jun Huang
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 negative-head uncertainty instead of positive-head uncertainty alone (ablation) | BUAL (B-LC) · CIFAR-10 (openness 0.6) | final accuracy (%): 87.5 -> 89.4 | Improved |
| combine positive and negative uncertainty (no balancing factors) (ablation) | BUAL (B-LC) · CIFAR-10 (openness 0.6) | final accuracy (%): 89.4 -> 90.8 | Improved |
| add global balancing factor r (ablation) | BUAL (B-LC) · CIFAR-10 (openness 0.6) | final accuracy (%): 90.8 -> 91.3 | Improved |
| add local balancing factor f_aux (ablation) | BUAL (B-LC) · CIFAR-10 (openness 0.6) | final accuracy (%): 91.3 -> 92.5 | Improved |
| add balancing factors and negative head on CIFAR-100 (ablation) | BUAL (B-LC) · CIFAR-100 (openness 0.6) | final accuracy (%): 63.5 -> 62.4 (w/o w) -> 67.5 | Mixed |
| negative learning fine-tuning + bidirectional uncertainty vs Margin sampling | Margin · CIFAR-10 (openness 0.6) | final accuracy (%): 89.0 -> 92.6 | Improved |
| open-set annotation methods LfOSA/CCAL when openness is low | LfOSA · CIFAR-10 (openness 0.2) | final accuracy (%): 83.3 -> 73.7 | Hurt |
| diversity-based selection (Coreset, BADGE) at high openness | Coreset · CIFAR-10 (openness 0.4) | final accuracy (%): 82.5 -> 81.8 | No clear effect |
Closest papers by meaning
- Rethinking Epistemic and Aleatoric Uncertainty for Active Open-Set Annotation: An Energy-Based Approach (CVPR 2025 · similarity 0.91)
- Entropic Open-Set Active Learning (AAAI 2024 · similarity 0.88)
- Inconsistency-Based Data-Centric Active Open-Set Annotation (AAAI 2024 · similarity 0.88)
- Towards Cost-Effective Learning: A Synergy of Semi-Supervised and Active Learning (CVPR 2025 · similarity 0.86)
- Uncertainty Herding: One Active Learning Method for All Label Budgets (ICLR 2025 · similarity 0.85)
- Active Learning with LLMs for Partially Observed and Cost-Aware Scenarios (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.
