Metraipolis › archive › dc_fed143d79867
Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges — arXiv:2609.01210
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.01210
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T06:33:59.865Z
- SHA-256 of the archived text
83008d5e05ce932bcfe4017bb12dbae1ea8f1bcc8a0e5925254c4fdbcf38da5c- Stable URL of this page
https://metraipolis.com/documento/dc_fed143d79867
How to cite
Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges — arXiv:2609.01210 — arXiv (en), https://arxiv.org/abs/2609.01210 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fed143d79867 (Metraipolis archive dc_fed143d79867, sha256 83008d5e05ce932bcfe4017bb12dbae1ea8f1bcc8a0e5925254c4fdbcf38da5c)
BibTeX
@misc{dcfed143d79867,
title = {Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges — arXiv:2609.01210},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.01210},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fed143d79867; sha256 83008d5e05ce932bcfe4017bb12dbae1ea8f1bcc8a0e5925254c4fdbcf38da5c},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.01210. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Safety benchmarks for large language models often assess the risk of a user query, although the outcome of question answering depends on whether the response violates a policy. This distinction is critical in Chinese harmful-content evaluation, where linguistic variation and adversarial transformations can obscure risky intent. We introduce C-SafeQA, a policy-grounded benchmark for response-level Chinese safety evaluation. It comprises 538 base queries and 8,877 adversarial queries answered by four full-model LLM deployments, yielding 37,660 query-response records labeled safe, unsafe, or disputed. Reference labels are generated through agreement-aware multi-model adjudication and blind audits of stratified subsets by three safety experts. C-SafeQA supports both evaluation of target-model safety and auditing of seven automated safety judges against shared reference labels. Unsafe-response rates range from 0.93% to 3.35% on base queries and from 11.68% to 30.05% on adversarial queries. On the adversarial subset, judges show substantial trade-offs between unsafe-response recall and risk-query-conditioned safe-response false positive rate, and no judge dominates all metrics. Both acrostic transformations reduce unsafe recall for all seven judges, revealing mechanism-specific evaluator weaknesses. Dataset records, metadata, verification code, and judge scripts are publicly released to support recomputation, while benchmark construction, target-response generation, and private adjudication remain outside the release boundary. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-02 in cs.AI). Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.01210 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-17T06:15:12.869Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).