{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,13]],"date-time":"2026-10-13T00:50:28Z","timestamp":1791852628409,"version":"4.3.4"},"posted":{"date-parts":[[2026]]},"group-title":"SSRN","reference-count":26,"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>Large language model (LLM) agents increasingly generate candidate trading signals, but many \"alpha mining\" pipelines reward the agent on in-sample return quality alone, leaving transaction costs to be discovered at implementation. We build a closed-loop system in which a locally-hosted LLM (Qwen2.5-Coder-14B-Instruct) proposes daily cross-sectional equity factors in an AST-validated expression language; each candidate is backtested on a reconstructed point-in-time S&amp;P universe (1999-2026, not a survivorship-biased snapshot) and priced with an explicit spread-and-impact transaction-cost model; diagnostics feed back to the agent every round. We isolate the effect of an execution-cost penalty in the search objective using matched, independently seeded re-optimizations: 21 five-year-train\/one-year-test walk-forward folds (2006-2026) crossed with four seed alphas from distinct predictor categories (reversal, volume, volatility, momentum) and three LLM seeds, for 12 re-optimized comparisons. An agent rewarded purely on gross return converges to 49-101%\/day turnover in every category and produces a -100% cumulative return (net Sharpe -4.7 to -11.2) despite positive average gross Sharpe (0.67); a cluster bootstrap over the 252 underlying fold-level comparisons (paired execution-aware minus baseline) puts the Net Sharpe improvement at +9.96 (95% CI [8.3, 11.6]). Rewarding the identical search net of cost cuts turnover 22-38x, drives near-universal adoption of long-lookback smoothing operators, and, for the volatility-anomaly seed, produces a net-profitable factor (Sharpe +0.35, CAGR ~3.8%). The other three categories remain net-negative, by substantially smaller margins. These results identify the objective specification of an automated search as an economically consequential research choice in its own right.\u200b<\/jats:p>","DOI":"10.2139\/ssrn.7498983","type":"posted-content","created":{"date-parts":[[2026,9,21]],"date-time":"2026-09-21T14:54:17Z","timestamp":1790002457000},"source":"Crossref","is-referenced-by-count":0,"title":["Execution-Aware Alpha Mining: Teaching LLM Factor Agents to Account for Trading Costs\u200b"],"prefix":"10.2139","author":[{"ORCID":"https:\/\/orcid.org\/0009-0002-2203-3367","authenticated-orcid":true,"given":"Raghuram","family":"Nagireddy","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","reference":[{"key":"ref1","doi-asserted-by":"crossref","first-page":"259","DOI":"10.1111\/j.1540-6261.2006.00836.x","article-title":"The cross-section of volatility and expected returns","volume":"61","author":"A Ang","year":"2006","journal-title":"Journal of Finance"},{"key":"ref2","doi-asserted-by":"crossref","first-page":"458","DOI":"10.1090\/noti1105","article-title":"Pseudo-mathematics and financial charlatanism: The effects of backtest overfitting on out-of-sample performance","volume":"61","author":"D H Bailey","year":"2014","journal-title":"Notices of the American Mathematical Society"},{"key":"ref3","author":"L Cao","year":"2025","journal-title":"Chain-of-Alpha: Unleashing the power of large language models for alpha mining in quantitative trading"},{"key":"ref4","doi-asserted-by":"crossref","first-page":"57","DOI":"10.1111\/j.1540-6261.1997.tb03808.x","article-title":"On persistence in mutual fund performance","volume":"52","author":"M M Carhart","year":"1997","journal-title":"Journal of Finance"},{"key":"ref5","doi-asserted-by":"crossref","DOI":"10.1016\/j.frl.2025.108923","article-title":"Multifaceted variability in LLM-driven stock recommendations","volume":"86","author":"S Chon","year":"2025","journal-title":"Finance Research Letters"},{"key":"ref6","doi-asserted-by":"crossref","first-page":"719","DOI":"10.1111\/j.1540-6261.2012.01729.x","article-title":"A simple way to estimate bid-ask spreads from daily high and low prices","volume":"67","author":"S A Corwin","year":"2012","journal-title":"Journal of Finance"},{"key":"ref7","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1016\/j.jfineco.2014.10.010","article-title":"A five-factor asset pricing model","volume":"116","author":"E F Fama","year":"2015","journal-title":"Journal of Financial Economics"},{"key":"ref8","author":"K R French","year":"2026","journal-title":"Data library: Fama\/French 5 factors (2x3) daily and momentum factor daily"},{"key":"ref9","doi-asserted-by":"crossref","first-page":"877","DOI":"10.1111\/0022-1082.00349","article-title":"The high-volume return premium","volume":"56","author":"S Gervais","year":"2001","journal-title":"Journal of Finance"},{"key":"ref10","author":"R C Grinold","year":"2000","journal-title":"Active Portfolio Management"},{"key":"ref11","author":"J Han","year":"2026","journal-title":"QuantaAlpha: An evolutionary framework for LLM-driven alpha mining"},{"key":"ref12","doi-asserted-by":"crossref","first-page":"881","DOI":"10.1111\/j.1540-6261.1990.tb05110.x","article-title":"Evidence of predictable behavior of security returns","volume":"45","author":"N Jegadeesh","year":"1990","journal-title":"Journal of Finance"},{"key":"ref13","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":"Journal of Finance"},{"key":"ref14","first-page":"72","author":"Z Kakushadze","year":"2016","journal-title":"101 formulaic alphas"},{"key":"ref15","doi-asserted-by":"crossref","first-page":"1039","DOI":"10.1111\/j.1540-6261.2004.00656.x","article-title":"Are momentum profits robust to trading costs","volume":"59","author":"R A Korajczyk","year":"2004","journal-title":"Journal of Finance"},{"key":"ref16","doi-asserted-by":"crossref","first-page":"11715","DOI":"10.18653\/v1\/2026.acl-long.538","article-title":"Cognitive alpha mining via LLM-driven code-based evolution","volume":"1","author":"F Liu","year":"2026","journal-title":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics"},{"key":"ref17","author":"Data Nasdaq","year":"2026","journal-title":"Sharadar core us equities bundle: SEP daily prices and SP500 constituent history"},{"key":"ref18","doi-asserted-by":"crossref","first-page":"703","DOI":"10.2307\/1913610","article-title":"A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix","volume":"55","author":"W K Newey","year":"1987","journal-title":"Econometrica"},{"key":"ref19","doi-asserted-by":"crossref","first-page":"104","DOI":"10.1093\/rfs\/hhv063","article-title":"A taxonomy of anomalies and their trading costs","volume":"29","author":"R Novy-Marx","year":"2016","journal-title":"Review of Financial Studies"},{"key":"ref20","author":"M L\u00f3pez De Prado","year":"2018","journal-title":"Advances in Financial Machine Learning"},{"key":"ref21","doi-asserted-by":"crossref","first-page":"997","DOI":"10.1609\/aaai.v40i2.37069","article-title":"Navigating the alpha jungle: An LLM-powered MCTS framework for formulaic alpha factor mining","author":"Y Shi","year":"2026","journal-title":"Proceedings of the AAAI Conference on Artificial Intelligence"},{"key":"ref22","first-page":"2813","article-title":"AlphaAgent: LLM-driven alpha mining with regularized exploration to counteract alpha decay","author":"Z Tang","year":"2025","journal-title":"Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining"},{"key":"ref23","author":"J Yi","year":"2026","journal-title":"AlphaSchema: Exploring the space of trading semantics for LLM-based alpha mining"},{"key":"ref24","author":"X Yu","year":"2026","journal-title":"Towards autonomous formulaic alpha discovery: An evolutionary computation perspective"},{"key":"ref25","author":"Y Zhao","year":"2026","journal-title":"AlphaPareto: Formulaic alpha discovery with LLM-guided multi-objective reinforcement learning"},{"key":"ref26","author":"A Appendix","journal-title":"Methodological details"}],"container-title":[],"original-title":[],"deposited":{"date-parts":[[2026,10,13]],"date-time":"2026-10-13T00:25:06Z","timestamp":1791851106000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ssrn.com\/abstract=7498983"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":26,"URL":"https:\/\/doi.org\/10.2139\/ssrn.7498983","relation":{},"subject":[],"published":{"date-parts":[[2026]]},"subtype":"preprint"}}