{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.01679", "kind": "paper", "url": "https://arxiv.org/abs/2608.01679"}], "authors": ["Qiuyang Zhan", "Rui Zhang", "Sheng Guo", "Lepeng Zhao", "Zhuotao Liu"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.01679", "source_url": "https://arxiv.org/abs/2608.01679"}, "categories": ["Agents", "General AI", "Language & Knowledge", "Safety", "Self-Evolution"], "description": {"en": "AuthMem-Bench evaluates authority collapse in persistent memory for LLM agents. It uses a paired benchmark holding claims and tasks fixed while varying source authority, measuring write-time collapse, authorization errors, and automatic authority preservation."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.01679", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.01679", "title": "AuthMem-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Tsinghua University", "East China Normal University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2608.01679", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2608.01679", "source_url": "https://openalex.org/W7172446784"}, "key": "claire-radar:2608.01679", "modality": null, "name": "AuthMem-Bench", "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_authmem-bench_56d03df5", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "53904908cb45f64cfafa501dc54f5cc38a761a79fe5ee25cd0cb916b564a5b65", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2608.01679"}, "publisher": null, "released": "2026-08-03", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.01679", "source_url": "https://arxiv.org/abs/2608.01679"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-01679", "source": "claire_radar", "source_benchmark_id": "2608.01679", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2608.01679", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "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": "59f3b5e566de1dbc5839708d614c163fef369422470391756f0e748c095c89a6", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "AuthMem-Bench evaluates authority collapse in persistent memory for LLM agents. It uses a paired benchmark holding claims and tasks fixed while varying source authority, measuring write-time collapse, authorization errors, and automatic authority preservation.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce AuthMem-Bench, a controlled paired benchmark that holds the focal claim and downstream task fixed while varying only source authority."}, "familyId": "bmf_3b53359afe7a", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_authmem-bench_56d03df5", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2608.01679", "paper": "https://arxiv.org/abs/2608.01679", "pdf": "https://arxiv.org/pdf/2608.01679", "project": null, "report": "https://arxiv.org/abs/2608.01679"}, "metrics": [], "motivation": "Persistent memory allows (self-evolving) LLM agents to adapt across tasks by consolidating heterogeneous interaction histories into reusable facts, preferences, observations, and rules.", "name": "AuthMem-Bench", "oneLine": "AuthMem-Bench evaluates authority collapse in persistent memory for LLM agents. It uses a paired benchmark holding claims and tasks fixed while varying source authority, measuring write-time collapse, authorization errors, and automatic authority preservation.", "paperTitle": "When Memory Becomes Authority: Benchmarking Authority Collapse at the Memory Consolidation Boundary", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2608.01679", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {"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}}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-08-03", "source": {"authors": ["Qiuyang Zhan", "Rui Zhang", "Sheng Guo", "Lepeng Zhao", "Zhuotao Liu"], "categories": ["cs.AI"], "id": "2608.01679", "title": "When Memory Becomes Authority: Benchmarking Authority Collapse at the Memory Consolidation Boundary", "type": "arxiv", "url": "https://arxiv.org/abs/2608.01679"}, "sourceUpdatedAt": "2026-08-04T08:28:14Z", "topics": ["Self-Evolution", "Agents", "Safety"], "venueAttempts": [], "whyItMatters": "Memory consolidation can erase authority constraints, leading to unauthorized actions. AuthMem-Bench provides a controlled benchmark to measure and improve authority preservation in memory systems, relevant for safe agent deployment."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
