Database Paper Browser

Back to papers

Fault Lines: Benchmarking the Impact of Label Data Quality on ML Robustness and Fairness

Summary: FAULT LINES: a model-agnostic benchmark (15 datasets, systematic diverse label corruptions) plus an evaluation suite to measure robustness and fairness across 22 SOTA classifiers. Key result: many models resist random noise but <10% biased noise causes large accuracy and fairness losses; transformers often handle biased noise better than GBDTs but with higher tuning-dependent variance. (summarized by gpt-5-mini on Mar 13 2026)

Paper ID
14349
Venue
VLDB
Year
2026
Pagerank
4.1945683e-05
Overall Rank
10,306 | 28.31%
DOI
10.14778/3785297.3785308

Incoming Non-self Citations Over Time

No non-self incoming citations found for this paper in this database.

Authors

Incoming Citations (Sorted by Pagerank)

Showing 0 of 0 citing papers.

Rank Citing Paper Year Venue Pagerank
Previous Page 1 / 1 Next

Outgoing Citations (Sorted by Pagerank)

Showing 3 of 3 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank Cited Paper Year Venue Pagerank
1,404 Responsible Data Management 2020 VLDB 0.00012174977
3,396 Automatic Data Repair: Are We Ready to Deploy? 2024 VLDB 7.1455126e-05
6,553 How do Categorical Duplicates Affect ML? A New Benchmark and Empirical Analyses 2024 VLDB 5.0157344e-05
Previous Page 1 / 1 Next

Semantically Similar Papers