AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Breaking Correlation Shift via Conditional Invariant Regularizer

Mingyang Yi, Ruoyu Wang, Jiacheng Sun, Zhenguo Li, Zhi-Ming Ma

ICLR 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Out-of-distribution generalization theory · 0 citations (OpenAlex)

Read the paper

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 changedWhereMeasuredEffect
Regularize with CSV using observable spurious attributes (RCSV)ResNet50 · CelebAAvg accuracy (%): 91.6 -> 92.6Improved
Regularize with CSV using observable spurious attributes (RCSV)ResNet50 · WaterbirdsAvg accuracy (%): 90.2 -> 90.5Improved
Regularize with CSV using observable spurious attributes (RCSV)BERT · MultiNLIAvg accuracy (%): 82.0 -> 82.6Improved
Regularize with CSV using observable spurious attributes (RCSV)BERT · CivilCommentsAvg accuracy (%): 82.2 -> 83.7Improved
Estimate and regularize CSV without observable spurious attributes (RCSV_U)ResNet50 · CelebAAvg accuracy (%): 85.5 -> 89.8Improved
Estimate and regularize CSV without observable spurious attributes (RCSV_U)ResNet50 · WaterbirdsAvg accuracy (%): 84.3 -> 89.1Improved
Estimate and regularize CSV without observable spurious attributes (RCSV_U)BERT · MultiNLIAvg accuracy (%): 79.5 -> 80.0Improved
Estimate and regularize CSV without observable spurious attributes (RCSV_U)BERT · CivilCommentsAvg accuracy (%): 83.5 -> 81.7Hurt
Use observable spurious attributes for CSV estimation instead of the unobservable estimatorRCSV/RCSV_U · CelebA, Waterbirds, MultiNLI, CivilCommentsAvg accuracy (%)Improved
Use the stated inverse-frequency reweighted sampling strategy when estimating CSV (ablation)RCSV and RCSV_U · OOD benchmarksOOD generalizationImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy