{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.13104", "kind": "paper", "url": "https://arxiv.org/abs/2606.13104"}, {"id": "gh:floating-reeds/authoritybench", "kind": "repo", "url": "https://github.com/floating-reeds/AuthorityBench"}], "authors": ["Aryan Khurana", "Aravind Ramana RN", "Dhruv Kumar"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.13104", "source_url": "https://arxiv.org/abs/2606.13104"}, "categories": ["General AI", "Safety & Trustworthiness", "cs.LG"], "description": {"en": "AuthorityBench evaluates how citation-based authority signals affect epistemic behavior in LLMs using 220,564 prompts across four domains, with a 2x2 factorial design crossing claim veracity and citation veracity."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.13104", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.13104", "title": "AuthorityBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Birla Institute of Technology and Science, Pilani"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.13104", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.13104", "source_url": "https://openalex.org/W7164517795"}, "key": "claire-radar:2606.13104", "modality": null, "name": "AuthorityBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_authoritybench_5d25d571", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "8e2df8405bdd1f6a4bf447ed49fbd964273bff68678ecb32127e61e62a8bd7e4", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.13104"}, "publisher": null, "released": "2026-06-11", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.13104", "source_url": "https://arxiv.org/abs/2606.13104"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-13104", "source": "claire_radar", "source_benchmark_id": "2606.13104", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Safety & Trustworthiness", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/floating-reeds/AuthorityBench", "githubScope": "benchmark_repo", "githubStars": 0, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2606.13104", "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": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "cc87a45f373a49769ef8548228e8198ada140e0c2bb0e4c064a8cf4fd98198d3", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "AuthorityBench evaluates how citation-based authority signals affect epistemic behavior in LLMs using 220,564 prompts across four domains, with a 2x2 factorial design crossing claim veracity and citation veracity.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce AuthorityBench, a 220,564-prompt multi-domain benchmark that isolates how citation-based authority signals influence epistemic behavior in LLMs."}, "familyId": "bmf_fba66cb2685f", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/floating-reeds/AuthorityBench", "url": "https://github.com/floating-reeds/AuthorityBench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/floating-reeds/AuthorityBench", "url": "https://huggingface.co/papers/2606.13104"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/floating-reeds/AuthorityBench", "url": "https://arxiv.org/abs/2606.13104"}], "scope": "benchmark_repo", "url": "https://github.com/floating-reeds/AuthorityBench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_authoritybench_5d25d571", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/floating-reeds/AuthorityBench", "data": null, "hfPaper": "https://huggingface.co/papers/2606.13104", "paper": "https://arxiv.org/abs/2606.13104", "pdf": "https://arxiv.org/pdf/2606.13104", "project": null, "report": "https://arxiv.org/abs/2606.13104"}, "metrics": [], "motivation": "Large language models are increasingly deployed in citation-augmented settings, yet the effect of citation presence on model behavior independent of factual content remains poorly understood.", "name": "AuthorityBench", "oneLine": "AuthorityBench evaluates how citation-based authority signals affect epistemic behavior in LLMs using 220,564 prompts across four domains, with a 2x2 factorial design crossing claim veracity and citation veracity.", "paperTitle": "Authority, Truth, and Citation Bias: A Large-Scale Multi-Domain Benchmark for Studying Epistemic Susceptibility in Large Language Models", "primaryDomain": "General AI", "publication": {"evidence": "10 pages, 5 figures. Accepted to AI4GOOD and EIML at ICML 2026", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2606.13104", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "AI4GOOD and EIML at ICML 2026", "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "floating-reeds", "organizationType": "community", "role": "benchmark-publisher", "sourceUrl": "https://github.com/floating-reeds/AuthorityBench"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-11", "source": {"authors": ["Aryan Khurana", "Aravind Ramana RN", "Dhruv Kumar"], "categories": ["cs.LG"], "id": "2606.13104", "title": "Authority, Truth, and Citation Bias: A Large-Scale Multi-Domain Benchmark for Studying Epistemic Susceptibility in Large Language Models", "type": "arxiv", "url": "https://arxiv.org/abs/2606.13104"}, "sourceUpdatedAt": "2026-06-11T09:33:03Z", "topics": ["cs.LG"], "venueAttempts": [{"decisionRaw": "10 pages, 5 figures. Accepted to AI4GOOD and EIML at ICML 2026", "evidence": [{"level": "author-claim", "observedAt": "2026-10-14T01:46:58.058199Z", "rawValue": "10 pages, 5 figures. Accepted to AI4GOOD and EIML at ICML 2026", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2606.13104"}], "reviewStatus": "accepted", "venueName": "AI4GOOD and EIML at ICML 2026"}], "whyItMatters": "The benchmark isolates citation presence from content to measure susceptibility to citation-induced hallucination, providing a standardized protocol for assessing reliability in citation-augmented settings."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
