Research map / Label-efficient and robust learning / Noisy labels
Label Noise Robustness of Conformal Prediction
Bat-Sheva Einbinder, Shai Feldman, Stephen Bates, Anastasios N. Angelopoulos, Asaf Gendler, Yaniv Romano
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 |
|---|---|---|---|
| calibrate conformal prediction with noisy single-annotator labels instead of clean labels | Vanilla conformal prediction with ResNet-18 · CIFAR-10H | coverage of clean labels: 90% -> approximately 93% | Improved |
| add uniform flip label noise (epsilon 0.05) to calibration | Conformal prediction with oracle classifier · Synthetic multi-class classification | coverage of clean labels: 90% -> around 94% | Improved |
| use rare-to-most-frequent class noise instead of uniform flip | Conformal prediction with neural network · Synthetic multi-class classification | coverage of clean labels | Hurt |
| use APS score instead of HPS score under label noise | Conformal prediction · Synthetic multi-class classification | robustness of coverage to label noise | Improved |
| add dispersive or symmetric noise to responses | CQR conformal intervals · Synthetic regression | coverage of clean responses | Improved |
| add biased or contractive noise to responses | CQR conformal intervals · Synthetic regression | coverage of clean responses | Hurt |
| fit predictive model on clean training data instead of noisy (ablation) | CQR conformal intervals with biased noise · Synthetic regression | coverage requirement | Improved |
| calibrate with noisy labels | CQR and residual magnitude conformal intervals · AVA aesthetic regression | interval width and coverage | Mixed |
| calibrate conformal risk control with noisy labels (fine and coarse) | Conformal risk control with TResNet · Multi-label CIFAR-100N | FNR on clean labels | No clear effect |
| calibrate conformal risk control with noisy annotations (beta 0.1) | Conformal risk control with PraNet · Polyp segmentation | FNR on clean labels | No clear effect |
| use distribution-free risk bounds on noisy calibration | Conformal risk control, smooth miscoverage · Adversarial synthetic regression | smooth miscoverage validity | Improved |
| calibrate Rolling RC with noisy depth labels | Rolling RC with LeReS · Depth estimation online | image miscoverage on clean labels | No clear effect |
Closest papers by meaning
- Exploring the Noise Robustness of Online Conformal Prediction (NeurIPS 2025 · similarity 0.92)
- Robust Yet Efficient Conformal Prediction Sets (ICML 2024 · similarity 0.90)
- Provably Robust Conformal Prediction with Improved Efficiency (ICLR 2024 · similarity 0.89)
- Robust Conformal Prediction with a Single Binary Certificate (ICLR 2025 · similarity 0.89)
- One Sample is Enough to Make Conformal Prediction Robust (NeurIPS 2025 · similarity 0.89)
- Verifiably Robust Conformal Prediction (NeurIPS 2024 · similarity 0.89)
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.
