{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.25535", "kind": "paper", "url": "https://arxiv.org/abs/2605.25535"}, {"id": "gh:yeonjun-in/permembench", "kind": "repo", "url": "https://github.com/yeonjun-in/PerMemBench"}], "authors": ["Yeonjun In", "Wonjoong Kim", "Sangwu Park", "Kanghoon Yoon", "Chanyoung Park"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.25535", "source_url": "https://arxiv.org/abs/2605.25535"}, "categories": ["General AI", "Language & Knowledge", "cs.AI"], "description": {"en": "PerMemBench evaluates personalized memory systems for LLM agents using multi-year, multi-domain interaction histories across 20 user personas, measuring memory retention accuracy."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.25535", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.25535", "title": "PerMemBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.25535", "modality": null, "name": "PerMemBench", "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_permembench_15a1b01c", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "c6318f3c0fe7e0748a417b21910e2b31a4f816d8fe008e07c25b23d1cc446ad0", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.25535"}, "publisher": null, "released": "2026-05-25", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.25535", "source_url": "https://arxiv.org/abs/2605.25535"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-25535", "source": "claire_radar", "source_benchmark_id": "2605.25535", "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": "https://github.com/yeonjun-in/PerMemBench", "githubScope": "benchmark_repo", "githubStars": 12, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 44, "hfPaperUrl": "https://huggingface.co/papers/2605.25535", "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", "lastSuccessfulDate": "2026-09-30", "state": "stale"}}, "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": "f7bcccbe7d923b8a2bbed3c622d775edc3b6607c3633cd1a06c98d82b39db8b0", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "PerMemBench evaluates personalized memory systems for LLM agents using multi-year, multi-domain interaction histories across 20 user personas, measuring memory retention accuracy.", "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 PerMemBench, the first benchmark for evaluating personalized memory systems, featuring multi year, multi domain interaction histories across diverse user personas."}, "familyId": "bmf_fb7852f323c5", "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/yeonjun-in/PerMemBench", "url": "https://github.com/yeonjun-in/PerMemBench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/yeonjun-in/PerMemBench", "url": "https://huggingface.co/papers/2605.25535"}], "scope": "benchmark_repo", "url": "https://github.com/yeonjun-in/PerMemBench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_permembench_15a1b01c", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/yeonjun-in/PerMemBench", "data": null, "hfPaper": "https://huggingface.co/papers/2605.25535", "paper": "https://arxiv.org/abs/2605.25535", "pdf": "https://arxiv.org/pdf/2605.25535", "project": null, "report": "https://arxiv.org/abs/2605.25535"}, "metrics": [], "motivation": "Existing large language model (LLM) based memory systems apply universal, static policies that overlook a fundamental reality: the contexts that are worth storing in memory are different across users.", "name": "PerMemBench", "oneLine": "PerMemBench evaluates personalized memory systems for LLM agents using multi-year, multi-domain interaction histories across 20 user personas, measuring memory retention accuracy.", "paperTitle": "Personalize-then-Store: Benchmarking and Learning Personalized Memory for Long-horizon Agents", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.25535", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "PerMemBench team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/yeonjun-in/PerMemBench"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-05-25", "source": {"authors": ["Yeonjun In", "Wonjoong Kim", "Sangwu Park", "Kanghoon Yoon", "Chanyoung Park"], "categories": ["cs.AI"], "id": "2605.25535", "title": "Personalize-then-Store: Benchmarking and Learning Personalized Memory for Long-horizon Agents", "type": "arxiv", "url": "https://arxiv.org/abs/2605.25535"}, "sourceUpdatedAt": "2026-05-25T07:48:33Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "Universal memory policies waste budget on transient interactions and fail to preserve critical context. PerMemBench enables evaluation of personalization, revealing that accurate gating remains an open challenge."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
