Open source by Santander AI Lab. A benchmark dataset collection for evaluating the robustness of machine learning and LLM systems: well-known datasets republished in stressed form — controlled, documented input degradation ("shocks": noise, missingness, ambiguity, formatting changes, contradictions) applied to a subset of the records. We publish the stressed versions so the community can freely reuse them.
Part of Santander AI open source · santander.com.
This is an evolving collection; more stressed datasets will be added over time. For now it contains the first one.
| Dataset | Source (mother dataset) | Version | Status | Link |
|---|---|---|---|---|
| Stressed German Credit Dataset | UCI German Credit | SGCD-v0.1 | Released | datasets/stressed_german_credit_dataset/ |
Each dataset is self-contained under datasets/<dataset_name>/, with its own README.md,
CHANGELOG.md, CITATION.cff, LICENSE, and a data/ folder holding the data and
metadata. The shocks/stresses applied are specific to each dataset and documented inside
its own folder.
This repository is dual-licensed:
- Data & documentation — CC BY 4.0 (
LICENSE). The community is free to use these datasets; each dataset uses the same license as its mother (source) dataset (collection default: CC BY 4.0), and the citation author is always Santander AI Lab (see each dataset'sCITATION.cffand the collection-levelCITATION.cff). - Source code — Apache License 2.0 (
LICENSE-CODE). Applies to the validation tooling under.github/scripts/and the CI workflows under.github/workflows/. - Always attribute both the original source dataset and Santander AI Lab.
To cite the collection, see CITATION.cff; to cite an individual dataset,
use the CITATION.cff inside its folder.
This collection is curated by Santander AI Lab; data files are regenerated from a
documented method, not hand-edited. We welcome issue reports (data problems,
attribution, documentation) and new-dataset proposals — see
CONTRIBUTING.md. For a privacy concern about any record, follow
SECURITY.md (report privately — do not open a public issue).
Please also read our Code of Conduct.
This is an open source project from the Santander AI Lab, provided "as is" under its license, without warranties or conditions of any kind. It is not an official Banco Santander product or service, carries no commitment of production support, and does not constitute financial, legal or professional advice.
"Santander" and its logo are registered trademarks of Banco Santander, S.A. The project license does not grant any right to use them beyond factual attribution.
If you believe you have found a security vulnerability, follow our security policy — do not open a public issue. You are responsible for assessing the suitability of this project for your use case and for keeping your own deployments up to date.