{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.28934", "kind": "paper", "url": "https://arxiv.org/abs/2607.28934"}, {"id": "gh:martinlukk/fairfund-bench", "kind": "repo", "url": "https://github.com/martinlukk/fairfund-bench"}], "authors": ["Martin Lukk"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.28934", "source_url": "https://arxiv.org/abs/2607.28934"}, "categories": ["General AI", "Safety & Trustworthiness", "cs.CL"], "description": {"en": "A benchmark that audits distributive bias in LLM resource allocation, varying task type, comparison context, and audit transparency across 600 financial-assistance requests."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.28934", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.28934", "title": "FairFund-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2607.28934", "modality": null, "name": "FairFund-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_fairfund-bench_51148b6a", "origin_source": "arxiv", "recognition_confidence": "0.75", "record_sha256": "72fea3d8f0b1304d0f3ce327364d070cbf6a7289b74170d8d74cbda7709e855a", "relation": "introduces", "review_model": "claude-code-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T16:37:47.209552Z", "source_url": "https://arxiv.org/abs/2607.28934"}, "publisher": null, "released": "2026-07-31", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.28934", "source_url": "https://arxiv.org/abs/2607.28934"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-28934", "source": "claire_radar", "source_benchmark_id": "2607.28934", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Safety & Trustworthiness", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Relevant to the LLM fairness auditing debate, and the code and data are public.", "score": 45}, "capabilities": [], "confidence": "Medium", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T16:37:47.209552Z", "inputHash": "8fc8e2b788b4e56a62c0541029bf1e9f173f4af4b8a74418d9b170b7bfcb2fe1", "model": "claude-code-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "FairFund-Bench: Evaluating Distributive Bias in LLM Resource Allocation", "canonicalNameSource": "paper_title", "decisionReason": "Formally named benchmark with a fixed request set, four explicit scoring criteria, application to 14 models, and public code and data.", "evaluationEvidence": {"reuse": "is publicly available, and can be readily adapted to other substantive domains", "scoring": "The benchmark scores models on four criteria (demographic bias, deservingness alignment, cross-task consistency, and cross-context consistency)", "task": "The benchmark comprises 600 requests for financial assistance created from human-authored templates"}, "model": "claude-code-default", "reviewedAt": "2026-09-24T16:37:47.209552Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A benchmark that audits distributive bias in LLM resource allocation, varying task type, comparison context, and audit transparency across 600 financial-assistance requests.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "evaluation protocol evidence", "public artifact URL"], "snippet": "We show that this disagreement can arise from differences in audit format and introduce FairFund-Bench, a benchmark that systematically varies key features of previous audit designs: the evaluation task (rating, ranking, or allocation), comparison context (single or multi-stimulus), and whether the audit is transparent or disguised."}, "familyId": "bmf_dc1a401e2540", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_fairfund-bench_51148b6a", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/martinlukk/fairfund-bench", "data": null, "paper": "https://arxiv.org/abs/2607.28934", "pdf": "https://arxiv.org/pdf/2607.28934", "project": null, "report": "https://arxiv.org/abs/2607.28934"}, "metrics": [], "motivation": "Large language models (LLMs) are increasingly involved in the distribution of scarce resources, raising concerns about biased allocations based on characteristics like race and gender.", "name": "FairFund-Bench", "oneLine": "A benchmark that audits distributive bias in LLM resource allocation, varying task type, comparison context, and audit transparency across 600 financial-assistance requests.", "paperTitle": "FairFund-Bench: Evaluating Distributive Bias in LLM Resource Allocation", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.28934", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "publishers": [{"name": "FairFund-Bench authors (martinlukk)", "organizationType": "community", "role": "benchmark-publisher", "sourceUrl": "https://github.com/martinlukk/fairfund-bench"}], "readiness": "Runnable", "recognitionConfidence": 0.75, "relation": "introduces", "releasedAt": "2026-07-31", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Martin Lukk (University of Toronto)"], "categories": ["cs.CL", "cs.AI", "cs.CY"], "id": "2607.28934", "title": "FairFund-Bench: Evaluating Distributive Bias in LLM Resource Allocation", "type": "arxiv", "url": "https://arxiv.org/abs/2607.28934"}, "sourceUpdatedAt": "2026-07-31T01:29:16Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "It reports that audit format can change the direction of measured bias. It gives four fixed scoring criteria so fairness results can be compared across models."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
