{"paper_id":"2505.11243","title":"A Set-Sequence Model for Time Series","authors":["Elliot L. Epstein","Apaar Sadhwani","Kay Giesecke"],"paper_date":"2025-05-16","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2505.11243v2","page_url":"https://thequant.space/flowcharts/a-set-sequence-model-for-time-series/","scores":{"empirical_rigor":6,"math_complexity":7,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes the Set-Sequence model, a neural architecture that learns cross‑sectional structure directly and combines a permutation‑invariant set module with a sequence module.","quote":"We propose Set-Sequence, a model that learns cross-sectional structure directly, enhancing expressivity and eliminating manual feature engineering.","summary":"Many prediction problems across science and engineering, especially in finance and economics, involve large cross-sections of individual time series, where each unit (e.g., a loan, stock, or customer) is driven by unit-level features and latent cross-sectional dynamics. While sequence models have…","strategy":"Set-Sequence"},"assumptions":{"status":"available","items":[{"assumption":"The transition matrix and dynamics of the latent factor λt are known in the synthetic contagion task.","type":"model","stated":"explicit","quote":"We assume that the transition matrix is known, and we also assume that the dynamics of λt, λt+1 = βλt + αNt are known"},{"assumption":"Portfolio weights are L1‑normalized to unit leverage each timestep.","type":"execution","stated":"explicit","quote":"We L1 normalize the portfolio weights at each timestep to maintain unit leverage (||wt||1 = 1)."},{"assumption":"Transaction costs are subtracted from daily returns using a linear cost model.","type":"execution","stated":"explicit","quote":"cost(wt, wt−1) = 0.0005 ∥wt −wt−1∥1 + 0.0001 ∥max(−wt, 0)∥1."},{"assumption":"Only the 500 assets with the largest market cap at the last quarter of the training period are retained for the equity experiment.","type":"data","stated":"explicit","quote":"We include only the 500 assets with the largest market cap at the last quarter of the training period, mimicking the S&P 500 index."},{"assumption":"Missing loan features are handled with a missing‑indicator method.","type":"data","stated":"explicit","quote":"We deal with missing data in other features with the missing indicator method"}]},"datasets":{"status":"available","items":[{"name":"CRSP stock dataset","provider":"CRSP","kind":"market-prices","access":"commercial","period":"8‑year training windows, tested on follo","universe":"36,600 stocks, filtered to 500 largest‑cap assets","quote":"... We include only the 500 assets with the largest market cap at the last quarter of the training period","known_id":"crsp"},{"name":"CoreLogic Loan‑Level Market Analytics Dataset","provider":"CoreLogic","kind":"alternative","access":"proprietary","period":"1988‑2024","universe":"Loans from top 4 ZIP codes in greater Los Angeles area (~117,523 loans)","quote":"We provide additional details about the CoreLogic Loan-Level Market Analytics Dataset. The dataset"},{"name":"Synthetic contagion task","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":"1000 units","quote":"synthetic task with 1000 units in total, the Set-Sequence model’s set summary in layer 3 and layer 5, learns the true joint effect λ0,t"}],"known":[{"id":"crsp","name":"CRSP","provider":"Center for Research in Security Prices","kind":"market-prices","access":"commercial","quote":"For the equities portfolio prediction task we use base data derived from Center for Research in Security Prices, LLC (CRSP).","mentions":3,"used_in":2}],"sample":{"start":"2002-01","end":"2021-12","market":"U.S. equities","asset_class":"equities","frequency":"daily"},"data_resolution":"daily"},"transaction_costs":{"status":"available","modeled":"yes","headline_net_of_costs":true,"detail":"cost(wt, wt−1) = 0.0005 ∥wt −wt−1∥1 + 0.0001 ∥max(−wt, 0)∥1; net return rnet,t = rt − cost(wt, wt−1)","quote":"cost(wt, wt−1) = 0.0005 ∥wt −wt−1∥1 + 0.0001 ∥max(−wt, 0)∥1. The return at time t, can then be expressed as rnet,t = rt −cost(wt, wt−1).","source":"front_matter","rule":"R3","candidates":{"front_matter":"yes","evidence_card":"yes","evidence_card_quoted":true},"card_detail":"cost(wt, wt−1) = 0.0005 ∥wt −wt−1∥1 + 0.0001 ∥max(−wt, 0)∥1; net return rnet,t = rt − cost(wt, wt−1)","card_quote":"cost(wt, wt−1) = 0.0005 ∥wt −wt−1∥1 + 0.0001 ∥max(−wt, 0)∥1. The return at time t, can then be expressed as rnet,t = rt −cost(wt, wt−1).","meaning":"costs are part of the reported headline results"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"Set-Sequence achieves an annualized Sharpe ratio of 4.82 ± 0.12, outperforming all sequence baselines (e.g., S4 at 3.94) and improving by 22 % over the second‑best baseline.","kind":"performance","quote":"The Set-Sequence model outperforms every sequence model baseline. We note a high annualized mean Sharpe ratio over the full period of 4.82, 22% higher than the second best sequence model (S4)"},{"finding":"Set-Sequence’s Sharpe ratio standard deviation across five random seeds is 0.12, lower than Transformer (0.52) and S4 (0.29).","kind":"performance","quote":"with a Sharpe ratio standard deviation over 5 random initialized training runs of 0.12, compared with 0.52 for the Transformer and 0.29 for S4"},{"finding":"Against a domain‑specific CNN‑Transformer, Set-Sequence improves Sharpe ratio by 42 %.","kind":"performance","quote":"the general Set-Sequence model outperforms the CNN-Transformer designed specifically for the equities task by 42 % in terms of the Sharpe ratio"},{"finding":"In the synthetic task, the set summary in layer five correlates 0.98 with the true joint effect λ0,t.","kind":"statistical","quote":"with a high correlation (r = 0.98) between the true joint effect and the set summary in layer five"},{"finding":"When training with a net‑Sharpe objective that accounts for transaction costs, Set-Sequence still outperforms baselines (e.g., net Sharpe 1.64 for LongConv vs 2.17 for S4).","kind":"performance","quote":"Table 9: Out of sample annualized performance metrics net transaction costs 2002 - 2016, trained with a net sharpe ratio objective."}],"reported_performance":{"strategy":"Set-Sequence","sharpe":4.82,"sharpe_benchmark":null,"annual_return_pct":13,"total_return_pct":null,"max_drawdown_pct":null},"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":28,"citations":{"cited_by_in_corpus":1,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":1},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T17:38:20Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}