Research map / Label-efficient and robust learning / Recognition and long-tail
Estimating and Explaining Model Performance When Both Covariates and Labels Shift
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SEES-c instead of KLIEP for estimating accuracy-gap performance | Gradient boosting · INSURANCE real shifts | accuracy-gap RMSE (%): 5.00 -> 1.70 | Helped |
| Use SEES-c instead of BBSE for estimating accuracy-gap performance | Gradient boosting · EMPLOY real shifts | accuracy-gap RMSE (%): 5.20 -> 2.90 | Helped |
| Use SEES-c instead of KLIEP for estimating accuracy-gap performance | Gradient boosting · INCOME real shifts | accuracy-gap RMSE (%): 3.40 -> 1.90 | Helped |
| Model joint label and feature shift with SEES-d rather than label shift with BBSE | COVID-19 case study · COVID-19 simulated age-and-label shift | weight-estimation MSE: 0.144 -> 0.002 | Helped |
| Model joint label and feature shift with SEES-d rather than covariate shift with KLIEP | COVID-19 case study · COVID-19 simulated age-and-label shift | weight-estimation MSE: 0.573 -> 0.002 | Helped |
| Estimate the shift using SEES-d rather than BBSE | COVID-19 case study · COVID-19 simulated age-and-label shift | accuracy-gap estimate (%): 26.3 -> 16.7; true gap 15.5 | Helped |
| Estimate the shift using SEES-d rather than KLIEP | COVID-19 case study · COVID-19 simulated age-and-label shift | accuracy-gap estimate (%): 1.58 -> 16.7; true gap 15.5 | Helped |
| Use SEES-c or SEES-d rather than BBSE or KLIEP as sample size increases | Accuracy-shift estimation under simulated 1-SJS · BANKCHURN, COVID-19, and CREDIT | squared L2 loss of accuracy estimation | Helped |
| Increase the number of jointly shifted features (ablation) | SEES-c and SEES-d · Simulated shifts with 10,000 samples | accuracy-estimation error | Hurt |
| Use SEES-c or SEES-d instead of BBSE under 0-SJS (label shift) (ablation) | Accuracy-shift estimation · 0-SJS / label shift | estimation error | Hurt |
Closest papers by meaning
- Sequential Harmful Shift Detection Without Labels (NeurIPS 2024 · similarity 0.87)
- Estimating Model Performance Under Covariate Shift Without Labels (NeurIPS 2025 · similarity 0.87)
- Beyond the Training Distribution: Evaluating Predictions Under Distribution Shift and Selection Bias (NeurIPS 2026 · similarity 0.87)
- "What is Different Between These Datasets?" A Framework for Explaining Data Distribution Shifts (NeurIPS 2026 · similarity 0.86)
- Optimal Aggregation of Prediction Intervals under Unsupervised Domain Shift (NeurIPS 2024 · similarity 0.86)
- Generalization vs Specialization under Concept Shift (NeurIPS 2025 · 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.