{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.20487", "kind": "paper", "url": "https://arxiv.org/abs/2606.20487"}], "authors": ["Shu Yao", "Yuhua Luo", "Qian Long", "Jingru Fan", "Yuheng Wang", "Lin Wu", "Zhuoyuan Yu", "Yufan Dang", "Huatao Li", "Chen Qian"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.20487", "source_url": "https://arxiv.org/abs/2606.20487"}, "categories": ["Agents", "Agents & Tool Use", "Mobile & Personal Computing"], "description": {"en": "HeraBench is a fault-injected benchmark for multi-device agent workflows on Linux and Android, evaluating hierarchical replanning under injected strategy- and device-level failures."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.20487", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.20487", "title": "HeraBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Shanghai Jiao Tong University", "Shanghai Innovation Institute", "Southeast University", "Tsinghua University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.20487", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.20487", "source_url": "https://openalex.org/W7165397510"}, "key": "claire-radar:2606.20487", "modality": null, "name": "HeraBench", "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_herabench_8a4ad4a5", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "94e8596079d6451531008eba58db31cd240acae1f92b27adf9ddd728648888c3", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.20487"}, "publisher": null, "released": "2026-06-18", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.20487", "source_url": "https://arxiv.org/abs/2606.20487"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-20487", "source": "claire_radar", "source_benchmark_id": "2606.20487", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Mobile & Personal Computing"], "area": "Agents & Tool Use", "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/2606.20487", "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": "2026-10-02", "state": "stale"}}, "status": "stale"}, "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": "33d4b5b3d23fa2335ee9396a6fb77b16e6957b669a600505ea0b05387a0f4197", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "HeraBench is a fault-injected benchmark for multi-device agent workflows on Linux and Android, evaluating hierarchical replanning under injected strategy- and device-level failures.", "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": "To evaluate this capability, we introduce \\textbf{HeraBench}, a fault-injected benchmark that constructs cross-device workflows over Linux and Android devices and injects strategy- and device-level failures."}, "familyId": "bmf_c80a435ff4de", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_herabench_8a4ad4a5", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Consumer Technology"], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.20487", "paper": "https://arxiv.org/abs/2606.20487", "pdf": "https://arxiv.org/pdf/2606.20487", "project": null, "report": "https://arxiv.org/abs/2606.20487"}, "metrics": [], "motivation": "Real-world computer-use tasks often span multiple applications and devices, requiring agents to coordinate heterogeneous environments under dynamic runtime failures.", "name": "HeraBench", "oneLine": "HeraBench is a fault-injected benchmark for multi-device agent workflows on Linux and Android, evaluating hierarchical replanning under injected strategy- and device-level failures.", "paperTitle": "Beyond Global Replanning: Hierarchical Recovery for Cross-Device Agent Systems", "primaryDomain": "Mobile & Personal Computing", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.20487", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-18", "source": {"authors": ["Shu Yao", "Yuhua Luo", "Qian Long", "Jingru Fan", "Zhuoyuan Yu", "Yuheng Wang", "Lin Wu", "Yufan Dang", "Huatao Li", "Chen Qian"], "categories": ["cs.CL"], "id": "2606.20487", "title": "Beyond Global Replanning: Hierarchical Recovery for Cross-Device Agent Systems", "type": "arxiv", "url": "https://arxiv.org/abs/2606.20487"}, "sourceUpdatedAt": "2026-06-18T17:04:17Z", "topics": ["Agents"], "venueAttempts": [], "whyItMatters": "Current multi-device agent benchmarks lack systematic fault injection to test recovery capabilities, making it difficult to compare hierarchical versus global replanning approaches."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
