AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Mitigating dataset harms requires stewardship: Lessons from 1000 papers

Kenneth L Peng, Arunesh Mathur, Arvind Narayanan

NeurIPS 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Fair and quality-aware face recognition · 8 citations (OpenAlex)

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 changedWhereMeasuredEffect
Retract the parent dataset without addressing remaining copies, derivatives, or modelsMS-Celeb-1M · 2020 research papersdocumented uses: 54Mixed
Retract the parent dataset without addressing remaining copies, derivatives, or modelsDukeMTMC · 2020 research papersdocumented uses: 73Mixed
Create derivatives that retain original images and remain available after parent retractionMS-Celeb-1M · 2020 research papersuses through derivative datasets or pre-trained models: 17; 26Mixed
Create derivatives that retain original images and remain available after parent retractionDukeMTMC · 2020 research papersuses through derivative datasets: 72 of 73Mixed
Make retraction reasons and requests not to use the data explicit and visibleDataset retraction · Retracted datasetscontinued useImproved
Make license restrictions explicit about whether they cover trained models and derivativesNon-commercial datasets · Production usecommercially licensed model releases: 18Improved
Use a non-commercial license without specifying its implications for pre-trained modelsMS-Celeb-1M · GitHub repositories containing pre-trained modelsrepositories including the same non-commercial designation: 3 of 21Hurt
Allow technological progress to make a research benchmark viable for production applicationsLFW · LFW website benchmark resultscommercial results: 3 of 38; 41 of 49Mixed
Use dataset derivatives to enable new applications, add sensitive annotations, or process imagesLFW, DukeMTMC, and MS-Celeb-1M derivatives · 41 identified derivativesderivatives with new applications: Twenty-one of 41Mixed
Rely on conventional citation practices rather than stable identifiers and dataset-specific trackingDataset research · Sampled papers and citation casescitations unable to be disambiguated: 42 of 446Hurt

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