{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.06055", "kind": "paper", "url": "https://arxiv.org/abs/2606.06055"}], "authors": ["Lingxiang Xu", "Jiaoyun Yang", "Min Hu", "Ning An"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.06055", "source_url": "https://arxiv.org/abs/2606.06055"}, "categories": ["General AI", "Language & Knowledge", "cs.AI"], "description": {"en": "HUSH-Bench evaluates conversational agents' use of sensitive history under a conservative policy, with 2,400 prompts and matched no-memory references, measuring unsolicited integration and memory access."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.06055", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.06055", "title": "HUSH-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Harvard University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.06055", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.06055", "source_url": "https://openalex.org/W7163706171"}, "key": "claire-radar:2606.06055", "modality": null, "name": "HUSH-Bench", "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_hush-bench_669096c5", "origin_source": "arxiv", "recognition_confidence": "0.85", "record_sha256": "42383d2f36df2bb8827b6807f77ce0c1048a097ec7fa3a49e48f6fed12a06d8a", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.06055"}, "publisher": null, "released": "2026-06-04", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.06055", "source_url": "https://arxiv.org/abs/2606.06055"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-06055", "source": "claire_radar", "source_benchmark_id": "2606.06055", "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": null, "hfPaperUrl": "https://huggingface.co/papers/2606.06055", "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", "state": "unavailable"}}, "status": "unavailable"}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "6c9b8b24af5d4c4d5ed74589f0eff744c9f7dcdeb6b5afb9d2eb55c49067053c", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "HUSH-Bench evaluates conversational agents' use of sensitive history under a conservative policy, with 2,400 prompts and matched no-memory references, measuring unsolicited integration and memory access.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence"], "snippet": "We introduce HUSH-Bench, a controlled benchmark of 2,400 benign prompts paired with histories containing one marked sensitive disclosure and matched no-memory references."}, "familyId": "bmf_9824d9a002a6", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_hush-bench_669096c5", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.06055", "paper": "https://arxiv.org/abs/2606.06055", "pdf": "https://arxiv.org/pdf/2606.06055", "project": null, "report": "https://arxiv.org/abs/2606.06055"}, "metrics": [], "motivation": "Long-term memory helps conversational agents maintain continuity across sessions, while relevance and current-turn warrant remain distinct decisions.", "name": "HUSH-Bench", "oneLine": "HUSH-Bench evaluates conversational agents' use of sensitive history under a conservative policy, with 2,400 prompts and matched no-memory references, measuring unsolicited integration and memory access.", "paperTitle": "HUSH-Bench: Measuring Memory-Use Boundaries for Sensitive History in Conversational Agents", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.06055", "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": "2026-06-04", "source": {"authors": ["Lingxiang Xu", "Jiaoyun Yang", "Min Hu", "Ning An"], "categories": ["cs.AI"], "id": "2606.06055", "title": "HUSH-Bench: Measuring Memory-Use Boundaries for Sensitive History in Conversational Agents", "type": "arxiv", "url": "https://arxiv.org/abs/2606.06055"}, "sourceUpdatedAt": "2026-08-03T14:34:11Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "It isolates memory retention from use in dialogue generation, showing that retrieval systems may surface sensitive data even without user request. This motivates separating storage, retrieval, and scope decisions in memory design."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
