{"paper":{"title":"PATRA: Pattern-Aware Alignment and Balanced Reasoning for Time Series Question Answering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"PATRA improves time series question answering by aligning extracted trends and seasonalities with language models while balancing rewards across task difficulties to support deeper reasoning.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Bin Yang, Chenjuan Guo, Christian S. Jensen, Junkai Lu, Peng Chen, Xingjian Wu, Yang Shu","submitted_at":"2026-02-26T16:20:03Z","abstract_excerpt":"Time series reasoning demands both the perception of complex dynamics and logical depth. However, existing LLM-based approaches exhibit two limitations: they often treat time series merely as text or images, failing to capture the patterns like trends and seasonalities needed to answer specific questions; and when trained on a mix of simple and complex tasks, simpler objectives often dominate the learning process, hindering the development of deep reasoning capabilities. To address these limitations, we propose the Pattern-Aware Alignment and Balanced Reasoning model (PATRA), introducing a pat"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"PATRA outperforms strong baselines across diverse Time Series Question Answering (TSQA) tasks, demonstrating superior cross-modal understanding and reasoning capability.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the pattern-aware mechanism successfully captures relevant dynamics without introducing noise and that the balanced reward truly enables deeper reasoning rather than just averaging performance across tasks.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"PATRA improves time series question answering by extracting patterns like trends and seasonalities for alignment and applying a task-aware balanced reward to support coherent reasoning.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"PATRA improves time series question answering by aligning extracted trends and seasonalities with language models while balancing rewards across task difficulties to support deeper reasoning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"cdde036bdf31ea32bdc831a18ab5a185a78b9d1d3b832ac6d81bb16ee846cb81"},"source":{"id":"2602.23161","kind":"arxiv","version":4},"verdict":{"id":"2d1755a5-dc67-46e2-a5f1-da37c2f01e97","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T19:01:50.499564Z","strongest_claim":"PATRA outperforms strong baselines across diverse Time Series Question Answering (TSQA) tasks, demonstrating superior cross-modal understanding and reasoning capability.","one_line_summary":"PATRA improves time series question answering by extracting patterns like trends and seasonalities for alignment and applying a task-aware balanced reward to support coherent reasoning.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the pattern-aware mechanism successfully captures relevant dynamics without introducing noise and that the balanced reward truly enables deeper reasoning rather than just averaging performance across tasks.","pith_extraction_headline":"PATRA improves time series question answering by aligning extracted trends and seasonalities with language models while balancing rewards across task difficulties to support deeper reasoning."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2602.23161/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"bdafd5ece19ce838e608d7a982f5a3dcfdae26061ee0406627ebd4f16665f498"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}