# SORRY-Bench

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

Measures the safety-refusal behavior of language models across a balanced taxonomy of unsafe requests and prompt mutations.

## Details

- Address: Online, Global
- Website: [sorry-bench.github.io/](<https://sorry-bench.github.io/>)
- Tags: Helpful, Science
- Category: Safety
- Status: Active
- Data licence: MIT
- Data access: gated
- Grader: llm-judge
- Organization: SORRY-Bench authors
- Released: 2024-06
- Paper: [arxiv.org/abs/2406.14598](<https://arxiv.org/abs/2406.14598>)
- Leaderboard: [sorry-bench.github.io/](<https://sorry-bench.github.io/>)
- Repository: [github.com/SORRY-Bench/sorry-bench](<https://github.com/SORRY-Bench/sorry-bench>)
- Task count: 450
- Top-score provenance: unknown
- ABI family: No

## Questions

### Where is SORRY-Bench?

SORRY-Bench is at Online, Global.

### How do I contact SORRY-Bench?

To contact SORRY-Bench, visit https://sorry-bench.github.io/.

## 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/sorry-bench>)
- [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.