{"paper":{"title":"SeesawNet: Towards Non-stationary Time Series Forecasting with Balanced Modeling of Common and Specific Dependencies","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"SeesawNet adaptively balances common and instance-specific dependencies in non-stationary time series forecasting.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hao Li, Liu Chong, Lu Zhang, Pengyang Wang, Yankai Chen, Yingjie Zhou","submitted_at":"2026-05-14T08:29:52Z","abstract_excerpt":"Instance normalization (IN) is widely used in non-stationary multivariate time series forecasting to reduce distribution shifts and highlight common patterns across samples. However, IN can over-smooth instance-specific structural information that is essential for modeling temporal and cross-channel heterogeneity. While prior methods further suppress distribution discrepancies or attempt to recover temporal specific dependencies, they often ignore a central tension: how to adaptively model common and instance-specific dependency based on each instance's non-stationary structures. To address th"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"SeesawNet consistently outperforms state-of-the-art methods on multiple real-world benchmarks by dynamically balancing common and instance-specific dependency modeling in both temporal and channel dimensions through ASNA.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the adaptive fusion of common dependencies from normalized sequences and specific dependencies from raw sequences, guided by instance-level non-stationarity, will reliably capture the required heterogeneity without introducing new smoothing artifacts or overfitting to training distribution shifts.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SeesawNet dynamically balances common and instance-specific dependencies via ASNA in temporal and channel dimensions, outperforming prior methods on non-stationary forecasting benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SeesawNet adaptively balances common and instance-specific dependencies in non-stationary time series forecasting.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"82db9a855d29f880dbd7de53fe7c44a05c07ceb440bd909f82df7db89d6016e0"},"source":{"id":"2605.14551","kind":"arxiv","version":1},"verdict":{"id":"638e2cbb-2f53-4312-9cb8-f656306de9a7","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T01:51:26.012761Z","strongest_claim":"SeesawNet consistently outperforms state-of-the-art methods on multiple real-world benchmarks by dynamically balancing common and instance-specific dependency modeling in both temporal and channel dimensions through ASNA.","one_line_summary":"SeesawNet dynamically balances common and instance-specific dependencies via ASNA in temporal and channel dimensions, outperforming prior methods on non-stationary forecasting benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the adaptive fusion of common dependencies from normalized sequences and specific dependencies from raw sequences, guided by instance-level non-stationarity, will reliably capture the required heterogeneity without introducing new smoothing artifacts or overfitting to training distribution shifts.","pith_extraction_headline":"SeesawNet adaptively balances common and instance-specific dependencies in non-stationary time series forecasting."},"references":{"count":21,"sample":[{"doi":"","year":2025,"title":"From dense to sparse: Event response for en- hanced residential load forecasting.IEEE Transactions on Instrumentation and Measurement,","work_id":"e8b8e84d-32a8-44c7-89d0-5a1c560330e5","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Pathformer: Multi-scale trans- formers with adaptive pathways for time series forecast- ing.arXiv preprint arXiv:2402.05956,","work_id":"aa5120c5-17cc-4a68-b04d-100ae5de9bdd","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Dish-ts: a general paradigm for alleviating distribution shift in time series forecasting","work_id":"4f385074-a836-4c03-b46b-e9e5915323e4","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Deep frequency derivative learning for non-stationary time series forecasting.arXiv preprint arXiv:2407.00502,","work_id":"f7a3436c-016d-4647-8984-a3ff88bfa95d","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Sin: Selective and interpretable normalization for long- term time series forecasting","work_id":"87b7b7bb-b192-4a69-a914-36daf83ab5df","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":21,"snapshot_sha256":"f7f3d98b3ee73b30e60c83fc45ec0a4aca23cd3a05b7178671c18cde32b9451e","internal_anchors":2},"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"}