Stress-Testing ML Pipelines with Adversarial Data Corruption
Summary: SAVAGE models realistic, interdependent data-quality failures as causal dependency graphs and uses bi-level black-box optimization to find worst-case corruptions. Experiments show that ~5% targeted errors can devastate ML pipelines, far beyond random corruption. (summarized by gpt-5.6-luna on Jul 24 2026)
Incoming Non-self Citations Over Time
No non-self incoming citations found for this paper in this database.
Authors
- 1. Jiongli Zhu (University of California San Diego)
- 2. Geyang Xu (University of California San Diego)
- 3. Felipe Lorenzi (University of California San Diego)
- 4. Boris Glavic (University of Illinois Chicago)
- 5. Babak Salimi (University of California San Diego)
BibTeX Citation
@article{zhu_vldb25,
title = {{Stress-Testing ML Pipelines with Adversarial Data Corruption}},
author = {Zhu, Jiongli and Xu, Geyang and Lorenzi, Felipe and Glavic, Boris and Salimi, Babak},
journal = {PVLDB},
series = {{VLDB} '25},
volume = {18},
number = {11},
pages = {4668--4681},
doi = {10.14778/3749646.3749721},
url = {https://doi.org/10.14778/3749646.3749721},
year = {2025}
}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 6 of 6 cited papers.
Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.
| Rank | Cited Paper | Year | Venue | Pagerank |
|---|---|---|---|---|
| 663 | A Formal Approach to Finding Explanations for Database Queries | 2014 | SIGMOD | 0.00015174751 |
| 1,951 | Interpretable Data-Based Explanations for Fairness Debugging | 2022 | SIGMOD | 9.4252389e-05 |
| 2,273 | SliceLine: Fast, Linear-Algebra-based Slice Finding for ML Model Debugging | 2021 | SIGMOD | 8.8230899e-05 |
| 5,291 | DiffPrep: Differentiable Data Preprocessing Pipeline Search for Learning over Tabular Data | 2023 | SIGMOD | 6.2801343e-05 |
| 6,709 | Through the Data Management Lens: Experimental Analysis and Evaluation of Fair Classification | 2022 | SIGMOD | 5.7964916e-05 |
| 6,884 | Explaining Inference Queries with Bayesian Optimization | 2021 | VLDB | 5.7465524e-05 |
Previous
Page 1 / 1
Next
Semantically Similar Papers
| # | Overall Rank | Paper | Year | Venue |
|---|---|---|---|---|
| 1 | 3,580 | Automatic Data Repair: Are We Ready to Deploy? | 2024 | VLDB |
| 2 | 7,490 | mlwhatif: What If You Could Stop Re-Implementing Your Machine Learning Pipeline Analyses Over and Over? | 2023 | VLDB |
| 3 | 8,050 | Capturing and Querying Fine-grained Provenance of Preprocessing Pipelines in Data Science | 2021 | VLDB |
| 4 | 6,637 | Causal Feature Selection for Algorithmic Fairness | 2022 | SIGMOD |
| 5 | 7,395 | Automating and Optimizing Data-Centric What-If Analyses on Native Machine Learning Pipelines | 2023 | SIGMOD |
| 6 | 9,245 | Towards Observability for Production Machine Learning Pipelines | 2022 | VLDB |
| 7 | 6,250 | Lightweight Inspection of Data Preprocessing in Native Machine Learning Pipelines | 2021 | CIDR |
| 8 | 7,232 | Saga: A Scalable Framework for Optimizing Data Cleaning Pipelines for Machine Learning Applications | 2023 | SIGMOD |
| 9 | 11,353 | Reconstructing and Querying ML Pipeline Intermediates | 2023 | CIDR |
| 10 | 10,604 | Fault Lines: Benchmarking the Impact of Label Data Quality on ML Robustness and Fairness | 2026 | VLDB |