{"record": {"aliases": [], "artifacts": [{"id": "gh:audn-ai/refusal-benchmark", "kind": "repo", "url": "https://github.com/audn-ai/refusal-benchmark"}, {"kind": "website", "url": "https://audn.ai/audncode"}], "categories": ["Finance", "Language & Knowledge", "cs.AI"], "description": {"en": "Measures LLM refusal rates on a set of harmful prompts via OpenAI-compatible endpoints."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://github.com/audn-ai/refusal-benchmark", "source": "claire_radar", "source_url": "https://github.com/audn-ai/refusal-benchmark", "title": "refusal-benchmark"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:github:audn-ai/refusal-benchmark", "modality": null, "name": "refusal-benchmark", "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-candidate", "display_eligible": "false", "origin_record_id": "bm_refusal-benchmark_83307c24", "origin_source": "github", "recognition_confidence": "0.4", "record_sha256": "0f0e94c66f5e1c4ad8b681edb14940a3b5dd6f400e44b3915db632c8e5713720", "relation": "unclear", "review_model": "deepseek-v4-pro", "review_state": "ai-name-audit-deferred", "reviewed_at": "2026-08-27T04:18:56.318436Z", "source_url": "https://github.com/audn-ai/refusal-benchmark"}, "publisher": null, "released": "2026-08-22", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:github:audn-ai/refusal-benchmark", "source_url": "https://github.com/audn-ai/refusal-benchmark"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-github-audn-ai-refusal-benchmark", "source": "claire_radar", "source_benchmark_id": "github:audn-ai/refusal-benchmark", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Finance"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/audn-ai/refusal-benchmark", "githubScope": "benchmark_repo", "githubStars": 46, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": null, "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"state": "not_applicable"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "The niche focus on abliterated models and lack of formal benchmark naming likely limits initial attention.", "score": 15}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-27T04:18:56.318436Z", "inputHash": "d8601bebde40d769b75ab948a189f6a5cf555935f35bf6a7a4190d9e57f02a63", "model": "deepseek-v4-pro"}, "curation": {"decisionReason": "The repository describes a measurement tool but does not formally declare a benchmark name or provide an independent release, making its status as a benchmark unclear.", "model": "deepseek-v4-pro", "reviewedAt": "2026-08-27T04:18:56.318436Z", "state": "ai-name-audit-deferred"}, "dataStatus": "primary-source-candidate", "demo": false, "description": "Measures LLM refusal rates on a set of harmful prompts via OpenAI-compatible endpoints.", "discoverySources": [{"id": "github:audn-ai/refusal-benchmark", "type": "github", "url": "https://github.com/audn-ai/refusal-benchmark"}], "displayEligible": false, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["discovered via github", "benchmark term in abstract", "evaluation protocol evidence", "public artifact URL", "no explicit benchmark release evidence"], "snippet": "refusal-benchmark 519+ harmful prompts to detect how abliterated AI models are # refusal-benchmark > 519+ harmful prompts to detect how abliterated AI models are."}, "familyId": "bmf_076cfc84d10b", "firstSeenAt": "2026-08-23", "heat": null, "id": "bm_refusal-benchmark_83307c24", "indexedAt": "2026-08-23T05:46:57.857309Z", "industrySectors": ["Financial Services"], "links": {"code": "https://github.com/audn-ai/refusal-benchmark", "project": "https://audn.ai/audncode"}, "metrics": [], "motivation": "refusal-benchmark 519+ harmful prompts to detect how abliterated AI models are # refusal-benchmark > 519+ harmful prompts to detect how abliterated AI models are.", "name": "refusal-benchmark", "oneLine": "Measures LLM refusal rates on a set of harmful prompts via OpenAI-compatible endpoints.", "paperTitle": "refusal-benchmark", "primaryDomain": "Finance", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://github.com/audn-ai/refusal-benchmark", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-23T05:46:57.857309Z"}, "publications": [], "ranking": {"90d": {"confidence": "Low", "coverage": 0.55, "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": 0.5139254941087447, "value": 46}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.55, "rank": 58, "score": 51}, "method": "attention-ranking-v13", "rank": 58, "score": 51}}, "readiness": "Runnable", "recognitionConfidence": 0.4, "relation": "unclear", "releasedAt": "2026-08-22", "source": {"authors": ["audn-ai"], "categories": [], "id": "github:audn-ai/refusal-benchmark", "title": "refusal-benchmark", "type": "github", "url": "https://github.com/audn-ai/refusal-benchmark"}, "sourceUpdatedAt": "2026-08-23T05:03:49Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "Provides a tool for safety evaluation by quantifying model compliance with harmful requests."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
