{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.24595", "kind": "paper", "url": "https://arxiv.org/abs/2606.24595"}], "authors": ["Enze Ma", "Yufan Zhou", "Wei-Chieh Huang", "Jie Yang", "Huanhuan Ma", "Zixuan Wang", "Chengze Li", "Chunyu Miao", "Philip S. Yu", "Zhen Wang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.24595", "source_url": "https://arxiv.org/abs/2606.24595"}, "categories": ["Agents", "General AI", "Language & Knowledge"], "description": {"en": "MEMPROBE evaluates long-term memory in LLM agents by reconstructing hidden user-state from agent memory across 50 simulated users with 31 hidden dimensions each."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.24595", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.24595", "title": "MEMPROBE"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["University of Illinois Chicago", "KU Leuven", "University of California San Diego"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.24595", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.24595", "source_url": "https://openalex.org/W7165807432"}, "key": "claire-radar:2606.24595", "modality": null, "name": "MEMPROBE", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "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_memprobe_9e043327", "origin_source": "arxiv", "recognition_confidence": "0.85", "record_sha256": "15eaf0ebb420b0af699dcc83d1982167dfab71858f45ceec5f8a9efe653414f2", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.24595"}, "publisher": null, "released": "2026-06-23", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.24595", "source_url": "https://arxiv.org/abs/2606.24595"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-24595", "source": "claire_radar", "source_benchmark_id": "2606.24595", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Machine Generated", "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": 3, "hfPaperUrl": "https://huggingface.co/papers/2606.24595", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-10-04", "state": "stale"}}, "status": "stale"}, "capabilities": [], "confidence": "Medium", "construction": "Original Synthetic", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "e3b0d2b5d0a4c211b19e81ca11157d54d040ed43f738876f5848b936945f3209", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "MEMPROBE evaluates long-term memory in LLM agents by reconstructing hidden user-state from agent memory across 50 simulated users with 31 hidden dimensions each.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "evaluation protocol evidence"], "snippet": "We instantiate this view in MEMPROBE, a benchmark in which a memory-equipped agent assists simulated users, each carrying a hidden, taxonomy-anchored user-state bank, across a trajectory of leak-controlled tasks, after which that bank is reconstructed from the agent's resulting memory under both full-store and top-k access."}, "familyId": "bmf_8b0112f7dc9d", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_memprobe_9e043327", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.24595", "paper": "https://arxiv.org/abs/2606.24595", "pdf": "https://arxiv.org/pdf/2606.24595", "project": null, "report": "https://arxiv.org/abs/2606.24595"}, "metrics": [], "motivation": "Long-term memory promises LLM agents that grow more capable across sessions, maintaining an accurate, evolving understanding of the user that interaction forms.", "name": "MEMPROBE", "oneLine": "MEMPROBE evaluates long-term memory in LLM agents by reconstructing hidden user-state from agent memory across 50 simulated users with 31 hidden dimensions each.", "paperTitle": "MEMPROBE: Probing Long-Term Agent Memory via Hidden User-State Recovery", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.24595", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 0.85, "relation": "introduces", "releasedAt": "2026-06-23", "source": {"authors": ["Enze Ma", "Yufan Zhou", "Wei-Chieh Huang", "Jie Yang", "Huanhuan Ma", "Zixuan Wang", "Chengze Li", "Chunyu Miao", "Philip S. Yu", "Zhen Wang"], "categories": ["cs.CL"], "id": "2606.24595", "title": "MEMPROBE: Probing Long-Term Agent Memory via Hidden User-State Recovery", "type": "arxiv", "url": "https://arxiv.org/abs/2606.24595"}, "sourceUpdatedAt": "2026-06-23T13:52:46Z", "topics": ["Agents"], "venueAttempts": [], "whyItMatters": "Offers a direct measure of memory fidelity as an auditable artifact, distinct from downstream task success, potentially improving agent memory evaluation."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
