{"record": {"aliases": ["Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges"], "artifacts": [{"id": "arxiv:2609.01210", "kind": "paper", "url": "https://arxiv.org/abs/2609.01210"}], "authors": ["Rui Yang", "Shuang Huang", "Junhua Liu", "Ziqi Zhao", "Qingzhong Yan", "Yuhang Sun", "Cong Liu", "Guoping Hu", "Rui Mei", "Jing Shao"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.01210", "source_url": "https://arxiv.org/abs/2609.01210"}, "categories": ["General AI", "Safety", "Safety & Trustworthiness"], "description": {"en": "A policy-grounded benchmark for response-level Chinese safety evaluation, assessing LLM responses and safety judges."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.01210", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.01210", "title": "C-SafeQA"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Shanghai Artificial Intelligence Laboratory", "Shanghai Innovation Institute"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2609.01210", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2609.01210", "source_url": "https://openalex.org/W7206152262"}, "key": "claire-radar:2609.01210", "modality": null, "name": "C-SafeQA", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "false", "origin_record_id": "bm_who-judges-the-judges-a-chinese-safety-qa-_0295f156", "origin_source": "arxiv", "recognition_confidence": "0.45", "record_sha256": "af625dffe69474365ae36e587940da9a3249c8338a84b2a691604f026d8e6022", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-name-audit-deferred", "reviewed_at": "2026-09-03T00:59:34.317835Z", "source_url": "https://arxiv.org/abs/2609.01210"}, "publisher": null, "released": "2026-09-01", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.01210", "source_url": "https://arxiv.org/abs/2609.01210"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-01210", "source": "claire_radar", "source_benchmark_id": "2609.01210", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Safety & Trustworthiness", "attention": {"asOf": "2026-09-02", "attemptedAt": "2026-09-02T06:30:32.214812Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2609.01210", "observedAt": "2026-09-02T06:30:32.214812Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-09-02T06:30:32.214812Z", "state": "fresh"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "Limited information beyond the abstract suggests modest initial attention.", "score": 25}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-03T00:59:34.317835Z", "inputHash": "b7a64ecc17ccd7a2f1ba47de8888eeeb149ffa48dcbe679e56f466f99a10c186", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "We introduce C-SafeQA, a policy-grounded benchmark for response-level Chinese safety evaluation.", "canonicalNameSource": "abstract", "decisionReason": "No artifactEvidence or project URL is supplied; a formal release and reusable path cannot be confirmed.", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-03T00:59:34.317835Z", "state": "ai-name-audit-deferred"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A policy-grounded benchmark for response-level Chinese safety evaluation, assessing LLM responses and safety judges.", "displayEligible": false, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence"], "snippet": "We introduce C-SafeQA, a policy-grounded benchmark for response-level Chinese safety evaluation."}, "familyId": "bmf_2195f11d4100", "firstSeenAt": "2026-09-02", "heat": null, "id": "bm_who-judges-the-judges-a-chinese-safety-qa-_0295f156", "indexedAt": "2026-09-02T05:21:06.631990Z", "industrySectors": [], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2609.01210", "pdf": "https://arxiv.org/pdf/2609.01210", "project": null, "report": "https://arxiv.org/abs/2609.01210"}, "metrics": [], "motivation": "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.", "name": "C-SafeQA", "oneLine": "A policy-grounded benchmark for response-level Chinese safety evaluation, assessing LLM responses and safety judges.", "paperTitle": "Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2609.01210", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-09-02T05:21:06.631990Z"}, "publications": [], "ranking": {"30d": {"confidence": "Low", "coverage": 0.3, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.0, "value": 0}}, "confidence": "Low", "coverage": 0.3, "rank": null, "score": 35}, "method": "attention-ranking-v13", "rank": null, "score": 35}, "90d": {"confidence": "Low", "coverage": 0.15, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.0, "value": 0}}, "confidence": "Low", "coverage": 0.15, "rank": null, "score": 43}, "method": "attention-ranking-v13", "rank": null, "score": 43}, "today": {"confidence": "Low", "coverage": 0.45, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": 2, "historyDate": "2026-08-31", "rank": null, "score": null, "windowDays": 1}, "level": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": 0.25, "percentile": 0.25, "value": 0}, "llmAttentionForecast": {"maxContribution": 0.25, "percentile": 0.72, "role": "bonus", "value": 72}}, "confidence": "Low", "coverage": 0.45, "rank": 7, "score": 44}, "method": "attention-ranking-v13", "rank": 7, "score": 44}}, "readiness": "Paper only", "recognitionConfidence": 0.45, "relation": "introduces", "releasedAt": "2026-09-01", "source": {"authors": ["Rui Yang", "Shuang Huang", "Junhua Liu", "Ziqi Zhao", "Qingzhong Yan", "Yuhang Sun", "Cong Liu", "Guoping Hu", "Rui Mei", "Jing Shao"], "categories": ["cs.CR", "cs.AI"], "id": "2609.01210", "title": "Who Judges the Judges? A Chinese Safety QA Benchmark for Evaluating LLM Responses and Safety Judges", "type": "arxiv", "url": "https://arxiv.org/abs/2609.01210"}, "sourceUpdatedAt": "2026-09-01T13:15:01Z", "topics": ["Safety"], "venueAttempts": [], "whyItMatters": "Targets a gap in Chinese safety evaluation and the evaluation of safety judges themselves."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
