Research map / Label-efficient and robust learning / Recognition and long-tail
Improving Confidence Estimates for Unfamiliar Examples
Zhizhong Li, Derek Hoiem
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 |
|---|---|---|---|
| Apply temperature scaling calibrated on familiar data | single model · Novel test splits across the evaluated image datasets | NLL | Improved |
| Use an ensemble with temperature-scaled members | ensemble · Novel test splits | NLL | Improved |
| Distill ensemble predictions on familiar labeled and general unlabeled examples | G-distillation · Novel test splits | NLL | Improved |
| Reduce predictions toward the familiar class prior according to novelty scores | NCR · Novel test splits | NLL | Mixed |
| Use G-distillation rather than standard distillation | distilled model · VOC-COCO | NLL | No clear effect |
| Use G-distillation rather than temperature scaling | single model · VOC-COCO novel split | NLL | Improved |
| Use G-distillation rather than temperature scaling | single model · VOC-COCO familiar split | NLL | Hurt |
| Add an ensemble of G-distilled models (ablation) | G-distillation · Novel test splits | NLL | Improved |
| Add an ensemble of G-distilled models instead of using a calibrated ensemble (ablation) | G-distillation · Novel test splits | NLL | Hurt |
Closest papers by meaning
- Enhanced Sample Selection with Confidence Tracking: Identifying Correctly Labeled Yet Hard-to-Learn Samples in Noisy Data (AAAI 2025 · similarity 0.85)
- Pearls from Pebbles: Improved Confidence Functions for Auto-labeling (NeurIPS 2024 · similarity 0.83)
- Quantifying Uncertainty in Error Consistency: Towards Reliable Behavioral Comparison of Classifiers (NeurIPS 2025 · similarity 0.83)
- Crossing the Validation crisis: Cross-validation reduces benchmarking variance surprisingly well (NeurIPS 2026 · similarity 0.83)
- Enhancing Sample Selection Against Label Noise by Cutting Mislabeled Easy Examples (NeurIPS 2025 · similarity 0.83)
- Your Pre-trained LLM is Secretly an Unsupervised Confidence Calibrator (NeurIPS 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.
