{"record": {"aliases": ["AMNESIA: A Large Scale Medical Unlearning Benchmark Suite with Disease-Informed Analysis"], "artifacts": [{"id": "arxiv:2605.30599", "kind": "paper", "url": "https://arxiv.org/abs/2605.30599"}], "categories": ["Biology & Drug Discovery", "Language & Knowledge", "cs.LG"], "description": {"en": "An open-source medical-unlearning benchmark containing 70,560 question-answer pairs from 8,820 patient notes across 11 disease categories, with factual-recall and clinical-reasoning evaluations."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.30599", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.30599", "title": "AMNESIA"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.30599", "modality": null, "name": "AMNESIA", "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_amnesia-a-large-scale-medical-unlearning-b_b95faf54", "origin_source": "arxiv", "recognition_confidence": "0.6", "record_sha256": "dabdef521b877921f9638cd883895779b1bf95a7da96f131705c399eb067d1c6", "relation": "aggregates", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T18:07:45.018840Z", "source_url": "https://arxiv.org/abs/2605.30599"}, "publisher": null, "released": "2026-05-28", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.30599", "source_url": "https://arxiv.org/abs/2605.30599"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-30599", "source": "claire_radar", "source_benchmark_id": "2605.30599", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["AMNESIA: A Large Scale Medical Unlearning Benchmark Suite with Disease-Informed Analysis"], "annotation": "Machine Generated", "applicationDomains": ["Biology & Drug Discovery"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "The combination of medical LLMs, unlearning, and a large open-source evaluation may attract near-term research interest.", "score": 6}, "capabilities": [], "confidence": "Medium", "construction": "Original Synthetic", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T18:07:45.018840Z", "inputHash": "b31ce312d057fd75641682103d2b4b484a654f8f8ba3f7e673c479a567226898", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "AMNESIA: A Large Scale Medical Unlearning Benchmark Suite with Disease-Informed Analysis", "canonicalNameSource": "paper_title", "decisionReason": "The supplied abstract formally names the benchmark, specifies its evaluation tasks and leakage metric, and identifies it as open source.", "evaluationEvidence": {"reuse": "the first large-scale, open source benchmark for medical unlearning", "scoring": "introduce a new metric for detecting leakage of medical terminology.", "task": "AMNESIA includes both factual questions testing direct recall and reasoning questions testing clinical inference."}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T18:07:45.018840Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "An open-source medical-unlearning benchmark containing 70,560 question-answer pairs from 8,820 patient notes across 11 disease categories, with factual-recall and clinical-reasoning evaluations.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce AMNESIA, the first large-scale, open source benchmark for medical unlearning, with 70,560 question-answer pairs from 8,820 patient notes across 11 disease categories."}, "familyId": "bmf_558e5eac4699", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_amnesia-a-large-scale-medical-unlearning-b_b95faf54", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Pharma & Biotech"], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2605.30599", "pdf": "https://arxiv.org/pdf/2605.30599", "project": null, "report": "https://arxiv.org/abs/2605.30599"}, "metrics": [], "motivation": "Medical knowledge is continuously evolving.", "name": "AMNESIA", "oneLine": "An open-source medical-unlearning benchmark containing 70,560 question-answer pairs from 8,820 patient notes across 11 disease categories, with factual-recall and clinical-reasoning evaluations.", "paperTitle": "AMNESIA: A Large Scale Medical Unlearning Benchmark Suite with Disease-Informed Analysis", "primaryDomain": "Biology & Drug Discovery", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.30599", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Paper only", "recognitionConfidence": 0.6, "relation": "aggregates", "releasedAt": "2026-05-28", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Saeedeh Davoudi", "Reihaneh Iranmanesh", "Ophir Frieder", "Nazli Goharian"], "categories": ["cs.LG", "cs.CL"], "id": "2605.30599", "title": "AMNESIA: A Large Scale Medical Unlearning Benchmark Suite with Disease-Informed Analysis", "type": "arxiv", "url": "https://arxiv.org/abs/2605.30599"}, "sourceUpdatedAt": "2026-05-28T21:46:46Z", "topics": ["cs.LG"], "venueAttempts": [], "whyItMatters": "It provides a reusable evaluation of whether unlearning methods remove patient information while preserving shared clinical knowledge."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
