{"paper":{"title":"AegisTS: A Hierarchical Agent System with Reinforcement Learning for Multivariate Time Series Data Cleaning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A hierarchical reinforcement learning agent system can jointly optimize the processing order and method selection to clean multiple quality issues in multivariate time series data without needing ground truth.","cross_cats":[],"primary_cat":"cs.DB","authors_text":"Lu Chen, Mourad Khayati, Tianyi Li, Yuanyuan Yao, Yuhan Shi","submitted_at":"2026-05-06T13:31:37Z","abstract_excerpt":"Multivariate time series (MTS) are frequently affected by co-occurring quality issues, such as missing values, outliers, and constraint violations, which significantly undermine downstream analytics. Existing cleaning approaches fix only a limited set of such issues, making them ill-suited for scenarios where multiple quality problems arise simultaneously. Furthermore, these methods commonly depend on the availability of ground truth data or domain-specific rules, both of which are rarely accessible in real-world applications.\n  In this paper, we introduce AegisTS, an agent system with reinfor"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our experimental results show that AegisTS consistently outperforms existing methods, achieving up to 96% improvement in data cleaning quality and 27% improvement in downstream performance.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the proposed dual-stage reward mechanism, which couples upstream cleaning quality with downstream task performance, can reliably guide the hierarchical agents to optimal pipelines in the absence of ground truth data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"AegisTS uses a two-level RL agent architecture with a dual-stage reward to jointly optimize cleaning order and method selection for multivariate time series, delivering up to 96% better cleaning quality and 27% better downstream performance without ground truth.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A hierarchical reinforcement learning agent system can jointly optimize the processing order and method selection to clean multiple quality issues in multivariate time series data without needing ground truth.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4f5f82bb40717c7614ffcfde57febc050132e3c21505ced589cd2b9758e9dde1"},"source":{"id":"2605.04902","kind":"arxiv","version":3},"verdict":{"id":"d0cef379-a393-4ef4-9089-ba7c95790684","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T16:21:01.387848Z","strongest_claim":"Our experimental results show that AegisTS consistently outperforms existing methods, achieving up to 96% improvement in data cleaning quality and 27% improvement in downstream performance.","one_line_summary":"AegisTS uses a two-level RL agent architecture with a dual-stage reward to jointly optimize cleaning order and method selection for multivariate time series, delivering up to 96% better cleaning quality and 27% better downstream performance without ground truth.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the proposed dual-stage reward mechanism, which couples upstream cleaning quality with downstream task performance, can reliably guide the hierarchical agents to optimal pipelines in the absence of ground truth data.","pith_extraction_headline":"A hierarchical reinforcement learning agent system can jointly optimize the processing order and method selection to clean multiple quality issues in multivariate time series data without needing ground truth."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.04902/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T10:41:00.159384Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T22:01:23.702943Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T14:03:43.831112Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"4424964b1041675537bdecce4eecd371dd85fafd526b7a27c7179d95a6bf0089"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}