Research map / Learning theory and algorithms / Prediction and decision losses
Questioning the Coverage-Length Metric in Conformal Prediction: When Shorter Intervals Are Not Better
Yizhou Min, Yizhou Lu, Lanqi Li, Zhen Zhang, Jiaye Teng
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 |
|---|---|---|---|
| apply Prejudicial Trick (null set with prob 1-p, adjusted alpha otherwise) | VCP · Synthetic Gaussian-mixture noise (alpha=0.10, p=0.96) | interval length: 22.894 -> 22.614 | Improved |
| apply Prejudicial Trick with p=0.98 instead of 0.96 (ablation) | PT-VCP · Synthetic (alpha=0.10) | interval length: 22.614 -> 22.714 | Mixed |
| apply Prejudicial Trick | VCP · Synthetic (alpha=0.20, p=0.96) | interval length: 21.886 -> 21.255 | Improved |
| apply Prejudicial Trick | VCP · bike (regression, bias 10) | interval length: 20.46 -> 19.59 | Improved |
| apply Prejudicial Trick | VCP · star (regression, bias 5) | interval length: 10.14 -> 9.63 | Improved |
| apply Prejudicial Trick | VCP · facebook-2 (regression, bias 10) | interval length: 20.97 -> 21.01 | Hurt |
| apply Prejudicial Trick | VCP · 10 regression datasets | interval length: smaller length in 9 out of 10 cases | Improved |
| apply Prejudicial Trick | RAPS · ImageNet-val resnet18 (bias 40) | prediction set size: 304.02 -> 295.60 | Improved |
| apply Prejudicial Trick | RAPS · ImageNet-val inception (bias 40) | prediction set size: 304.10 -> 297.25 | Improved |
| apply Prejudicial Trick (adds randomness) | VCP · meps-19 | interval stability: 0.00 -> 1.26 | Hurt |
| apply Prejudicial Trick (adds randomness) | VCP · bike | interval stability: 0.00 -> 0.58 | Hurt |
| retrain local scale estimator across runs | Localized CP · bike | interval stability: 0.03 | No clear effect |
Closest papers by meaning
- Length Optimization in Conformal Prediction (NeurIPS 2024 · similarity 0.90)
- Parametric Scaling Law of Tuning Bias in Conformal Prediction (ICML 2025 · similarity 0.89)
- Leave-One-Out Stable Conformal Prediction (ICLR 2025 · similarity 0.89)
- Reliable Uncertainty Quantification in Machine Learning via Conformal Prediction (AAAI 2025 · similarity 0.88)
- Backward Conformal Prediction (NeurIPS 2025 · similarity 0.88)
- Exploring the Noise Robustness of Online Conformal Prediction (NeurIPS 2025 · similarity 0.88)
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.
