# XSTest

XSTest is an AI benchmark; this page records what it measures and traces its data, grader, and scores to primary sources.

Tests whether safety-tuned language models refuse harmless prompts that superficially resemble unsafe requests.

## Details

- Address: Online, Global
- Website: [github.com/paul-rottger/xstest](<https://github.com/paul-rottger/xstest>)
- Tags: Helpful, Science
- Category: Safety
- Status: Active
- Data licence: CC-BY-4.0
- Data access: open
- Grader: mixed
- Organization: XSTest authors
- Released: 2023-07
- Paper: [aclanthology.org/2024.naacl-long.301/](<https://aclanthology.org/2024.naacl-long.301/>)
- Repository: [github.com/paul-rottger/xstest](<https://github.com/paul-rottger/xstest>)
- Task count: 450
- Top-score provenance: unknown
- ABI family: No

## Questions

### Where is XSTest?

XSTest is at Online, Global.

### How do I contact XSTest?

To contact XSTest, visit https://github.com/paul-rottger/xstest.

## Source

From [Benchmark primary sources](https://aibenchmarkindex.com/about), license CC0-1.0, checked 2026-10-15.

Last updated October 15, 2026.

## Links

- This benchmark on the [web page](<https://aibenchmarkindex.com/global/online/xstest>)
- [Safety in Online](<https://aibenchmarkindex.com/global/online/category/safety.md>)
- [More benchmarks in Online](<https://aibenchmarkindex.com/global/online.md>)

> **Check before you compare.** Benchmark methods and leaderboards change. Follow the cited primary source before using a result in a decision.