{"paper":{"title":"PULSE: Generative Phase Evolution for Non-Stationary Time Series Forecasting","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A simple MLP equipped with physical phase evolution modeling matches or exceeds complex models on non-stationary time series forecasting.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Fei Wang, Hu Chen, Xinyu Chen, Yangyou Liu, Yuankai Wu, Zezhi Shao","submitted_at":"2026-05-16T03:54:18Z","abstract_excerpt":"Time series forecasting under non-stationarity faces a fundamental tension between capturing stable representations and adapting to distribution shifts. Existing methods implicitly rely on static historical assumptions, leading to a critical failure mode we term Phase Amnesia, where models become blind to the evolving global context. To resolve this, we formalize non-stationary dynamics through three physical hypotheses: wold decomposition, dynamical phase evolution, and heteroscedastic manifold generation. These principles inspire PULSE, a physics-informed, plug-and-play framework adopting a "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"PULSE enables a simple MLP backbone to achieve state-of-the-art or highly competitive performance across 12 real-world benchmarks. This validates that a correct physics-informed inductive bias is far more critical than raw architectural complexity for non-stationary forecasting.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The three physical hypotheses (Wold decomposition, dynamical phase evolution, and heteroscedastic manifold generation) provide a valid and useful formalization of non-stationary dynamics that directly translates into an effective forecasting architecture.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"PULSE formalizes non-stationary time series via three physical hypotheses and uses phase-anchored disentanglement plus a Phase Router to let a simple MLP reach competitive performance on 12 benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A simple MLP equipped with physical phase evolution modeling matches or exceeds complex models on non-stationary time series forecasting.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"742df9f8ea6826433e8f6c4cd3d2b7497bdc59eb089af518be0ebf4453b16ae6"},"source":{"id":"2605.16793","kind":"arxiv","version":1},"verdict":{"id":"3f0829c5-a69e-4ff9-bd4f-a67d18a2729c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T21:23:34.949570Z","strongest_claim":"PULSE enables a simple MLP backbone to achieve state-of-the-art or highly competitive performance across 12 real-world benchmarks. This validates that a correct physics-informed inductive bias is far more critical than raw architectural complexity for non-stationary forecasting.","one_line_summary":"PULSE formalizes non-stationary time series via three physical hypotheses and uses phase-anchored disentanglement plus a Phase Router to let a simple MLP reach competitive performance on 12 benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The three physical hypotheses (Wold decomposition, dynamical phase evolution, and heteroscedastic manifold generation) provide a valid and useful formalization of non-stationary dynamics that directly translates into an effective forecasting architecture.","pith_extraction_headline":"A simple MLP equipped with physical phase evolution modeling matches or exceeds complex models on non-stationary time series forecasting."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.16793/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T21:31:19.324456Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T21:31:00.859639Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T19:01:56.293232Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.429147Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"bdd4e00f8ae4c781f0bd5fa86934cd9110a30e7ff7516656fb409e7d4c3502fa"},"references":{"count":51,"sample":[{"doi":"","year":null,"title":"International conference on learning representations , year=","work_id":"4fc3e1aa-23e2-4198-b33b-77047c926b08","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"IEEE Transactions on Knowledge and Data Engineering , year=","work_id":"a65d099c-3473-44c4-9322-761cee0a1b42","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"The Thirteenth International Conference on Learning Representations , year=","work_id":"3e9b5d1f-7231-46fc-975a-9931d004c27b","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Proceedings of the 30th ACM international conference on information & knowledge management , pages=","work_id":"f85ca5ae-dbb0-4945-8e66-6561a0f676bd","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2016,"title":"Information geometry and its applications , author=. 2016 , publisher=","work_id":"3562d5c4-2244-41ca-8bb9-b4f16bc51533","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":51,"snapshot_sha256":"033a729fc32acd5265ea484f03abd07642be23cd6be2d7f9046dff7e5a23601e","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"cc0628d09ebc5f32d0348e3619c92373ea2cd34264ab357b8ab8efbef8d0b275"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}