{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,20]],"date-time":"2026-10-20T14:40:48Z","timestamp":1792507248143,"version":"4.5.0"},"posted":{"date-parts":[[2026]]},"group-title":"SSRN","reference-count":36,"publisher":"Elsevier BV","license":[{"start":{"date-parts":[[2026,1,1]],"date-time":"2026-01-01T00:00:00Z","timestamp":1767225600000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/www.uspto.gov\/ip-policy\/copyright-policy\/copyright-basics"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"<jats:p>Deep forecasters for stock returns typically stack many task\u2010specific layers, which encourages overfitting on the small effective sample of financial regimes, while spending computation uniformly on calm and turbulent markets alike. We propose FinLoop, a compact looping transformer for financial signal prediction that iteratively refines a shared latent representation of recent price history. A single pre\u2010norm attention block is executed K times with input injection, positional re\u2010injection, per\u2010loop embeddings, and gated per\u2010loop LayerScale, which lifts the expressiveness ceiling of plain weight sharing while keeping early iterations nearly identity to resist oversmoothing on noisy financial data. Each loop emits an intermediate forecast, and an adaptive\u2010halting unit in the spirit of adaptive computation time weighs these forecasts into a coarse\u2010to\u2010fine ensemble and enables per\u2010sample early exit at inference. A rank\u2010aware objective couples magnitude calibration with pairwise ordering supervision, per\u2010loop deep supervision, and an expected\u2010depth penalty. On a transparent CSI\u2010300 protocol with 300 stocks (2016 to 2020), four market regimes, eight deep baselines, and a top\u2010k portfolio simulation, FinLoop attains the best information coefficients, Sharpe ratio, and shallowest maximum drawdown. On a NASDAQ\u2010100 corpus built from the identical protocol, whose test window is dominated by extreme regimes, FinLoop again attains the best information and rank coefficients.<\/jats:p>","DOI":"10.2139\/ssrn.7641919","type":"posted-content","created":{"date-parts":[[2026,10,20]],"date-time":"2026-10-20T13:50:33Z","timestamp":1792504233000},"source":"Crossref","is-referenced-by-count":0,"title":["FinLoop: A Weight-Shared Looping Transformer with Adaptive Halting for Financial Signal Prediction"],"prefix":"10.2139","author":[{"given":"Zongwei","family":"Jia","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","reference":[{"key":"ref1","doi-asserted-by":"crossref","first-page":"65","DOI":"10.1111\/j.1540-6261.1993.tb04702.x","article-title":"Returns to buying winners and selling losers: Implications for stock market efficiency","volume":"48","author":"N Jegadeesh","year":"1993","journal-title":"The Journal of Finance"},{"key":"ref2","doi-asserted-by":"crossref","first-page":"27","DOI":"10.1016\/0304-405X(88)90021-9","article-title":"Mean reversion in stock prices: Evidence and implications","volume":"22","author":"J Poterba","year":"1988","journal-title":"Journal of Financial Economics"},{"key":"ref3","author":"G Box","year":"2015","journal-title":"Time series analysis: Forecasting and control"},{"key":"ref4","first-page":"785","article-title":"Xgboost: A scalable tree boosting system","author":"T Chen","year":"2016","journal-title":"Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining"},{"key":"ref5","article-title":"Lightgbm: A highly efficient gradient boosting decision tree","volume":"30","author":"G Ke","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref6","author":"J Zou","year":"2022","journal-title":"Stock market prediction via deep learning techniques: A survey"},{"key":"ref7","doi-asserted-by":"crossref","first-page":"1735","DOI":"10.1162\/neco.1997.9.8.1735","article-title":"Long short-term memory","volume":"9","author":"S Hochreiter","year":"1997","journal-title":"Neural Computation"},{"key":"ref8","author":"J Chung","year":"2014","journal-title":"Empirical evaluation of gated recurrent neural networks on sequence modeling"},{"key":"ref9","first-page":"2627","article-title":"A dual-stage attention-based recurrent neural network for time series prediction","author":"Y Qin","year":"2017","journal-title":"Proceedings of the 26th International Joint Conference on Artificial Intelligence"},{"key":"ref10","author":"S Bai","year":"2018","journal-title":"An empirical evaluation of generic convolutional and recurrent networks for sequence modeling"},{"key":"ref11","article-title":"Attention is all you need","volume":"30","author":"A Vaswani","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref12","article-title":"A time series is worth 64 words: Long-term forecasting with transformers","author":"Y Nie","year":"2023","journal-title":"International Conference on Learning Representations"},{"key":"ref13","article-title":"Crossformer: Transformer utilizing cross-dimension dependency for multivariate time series forecasting","author":"Y Zhang","year":"2023","journal-title":"International Conference on Learning Representations"},{"key":"ref14","author":"Y Liu","year":"2024","journal-title":"itransformer: Inverted transformers are effective for time series forecasting"},{"key":"ref15","article-title":"Decomposable multiscale mixing for time series forecasting","author":"S Wang","year":"2024","journal-title":"International Conference on Learning Representations"},{"key":"ref16","article-title":"Timesnet: Temporal 2d-variation modeling for general time series analysis","author":"H Wu","year":"2023","journal-title":"International Conference on Learning Representations"},{"key":"ref17","article-title":"Segrnn: Segment recurrent neural network for long-term time series forecasting","author":"S Lin","year":"2025","journal-title":"IEEE Internet of Things Journal"},{"key":"ref18","article-title":"Timebridge: Non-stationarity matters for long-term time series forecasting","author":"P Liu","year":"2025","journal-title":"Proceedings of the 42nd International Conference on Machine Learning"},{"key":"ref19","author":"A Gu","year":"2023","journal-title":"Linear-time sequence modeling with selective state spaces"},{"key":"ref20","article-title":"Reversible instance normalization for accurate time-series forecasting against distribution shift","author":"T Kim","year":"2021","journal-title":"International Conference on Learning Representations"},{"key":"ref21","author":"Finloop"},{"key":"ref22","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3309547","article-title":"Temporal relational ranking for stock prediction","volume":"37","author":"F Feng","year":"2019","journal-title":"ACM Transactions on Information Systems"},{"key":"ref23","doi-asserted-by":"crossref","first-page":"9187","DOI":"10.1609\/aaai.v38i8.28770","article-title":"Ci-sthpan: Pre-trained attention network for stock selection with channel-independent spatio-temporal hypergraph","volume":"38","author":"H Xia","year":"2024","journal-title":"Proceedings of the AAAI Conference on Artificial Intelligence"},{"key":"ref24","doi-asserted-by":"crossref","first-page":"162","DOI":"10.1609\/aaai.v38i1.27767","article-title":"Market-guided stock transformer for stock price forecasting","volume":"38","author":"T Li","year":"2024","journal-title":"Proceedings of the AAAI Conference on Artificial Intelligence"},{"key":"ref25","doi-asserted-by":"crossref","DOI":"10.1016\/j.neucom.2023.127033","article-title":"Predicting stock market trends with self-supervised learning","volume":"568","author":"Z Ying","year":"2024","journal-title":"Neurocomputing"},{"key":"ref26","doi-asserted-by":"crossref","first-page":"4468","DOI":"10.1609\/aaai.v36i4.20369","article-title":"Factorvae: A probabilistic dynamic factor model based on variational autoencoder for predicting crosssectional stock returns","volume":"36","author":"Y Duan","year":"2022","journal-title":"Proceedings of the AAAI Conference on Artificial Intelligence"},{"key":"ref27","first-page":"1","article-title":"Finrl: Deep reinforcement learning framework to automate trading in quantitative finance","author":"X Liu","year":"2021","journal-title":"Proceedings of the Second ACM International Conference on AI in Finance"},{"key":"ref28","author":"Y Hu","year":"2025","journal-title":"Fintsb: A comprehensive and practical benchmark for financial time series forecasting"},{"key":"ref29","article-title":"Universal transformers","author":"M Dehghani","year":"2019","journal-title":"International Conference on Learning Representations"},{"key":"ref30","article-title":"Albert: A lite bert for self-supervised learning of language representations","author":"Z Lan","year":"2020","journal-title":"International Conference on Learning Representations"},{"key":"ref31","first-page":"32","article-title":"Going deeper with image transformers","author":"H Touvron","year":"2021","journal-title":"IEEE\/CVF International Conference on Computer Vision"},{"key":"ref32","author":"A Graves","year":"2016","journal-title":"Adaptive computation time for recurrent neural networks"},{"key":"ref33","first-page":"14855","article-title":"Reasoning with latent thoughts: On the power of looped transformers","volume":"2025","author":"N Saunshi","year":"2025","journal-title":"International Conference on Learning Representations"},{"key":"ref34","first-page":"14502","article-title":"Looped transformers for length generalization","volume":"2025","author":"Y Fan","year":"2025","journal-title":"International Conference on Learning Representations"},{"key":"ref35","author":"B Wu","year":"2025","journal-title":"Parallel loop transformer for efficient test-time computation scaling"},{"key":"ref36","article-title":"Pytorch: An imperative style, high-performance deep learning library","volume":"32","author":"A Paszke","year":"2019","journal-title":"Advances in Neural Information Processing Systems"}],"container-title":[],"original-title":[],"deposited":{"date-parts":[[2026,10,20]],"date-time":"2026-10-20T13:52:18Z","timestamp":1792504338000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ssrn.com\/abstract=7641919"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":36,"URL":"https:\/\/doi.org\/10.2139\/ssrn.7641919","relation":{},"subject":[],"published":{"date-parts":[[2026]]},"subtype":"preprint"}}