MemGuard-Alpha: Detecting and Filtering Memorization-Contaminated Signals in LLM-Based Financial Forecasting via Membership Inference and Cross-Model Disagreement

View PDF

Abstract:Large language models (LLMs) are increasingly used to generate financial alpha signals, yet growing evidence shows that LLMs memorize historical financial data from their training corpora, producing spurious predictive accuracy that collapses out-of-sample. This memorization-induced look-ahead bias threatens the validity of LLM-based quantitative strategies. Prior remedies -- model retraining and input anonymization -- are either prohibitively expensive or introduce significant information loss. No existing method offers practical, zero-cost signal-level filtering for real-time trading. We introduce MemGuard-Alpha, a post-generation framework comprising two algorithms: (i) the MemGuard Composite Score (MCS), which combines five membership inference attack (MIA) methods with temporal proximity features via logistic regression, achieving Cohen's d = 18.57 for contamination separation (d = 0.39-1.37 using MIA features alone); and (ii) Cross-Model Memorization Disagreement (CMMD), which exploits variation in training cutoff dates across LLMs to separate memorized signals from genuine reasoning. Evaluated across seven LLMs (124M-7B parameters), 50 S&P 100 stocks, 42,800 prompts, and five MIA methods over 5.5 years (2019-2024), CMMD achieves a Sharpe ratio of 4.11 versus 2.76 for unfiltered signals (49% improvement). Clean signals produce 14.48 bps average daily return versus 2.13 bps for tainted signals (7x difference). A striking crossover pattern emerges: in-sample accuracy rises with contamination (40.8% to 52.5%) while out-of-sample accuracy falls (47% to 42%), providing direct evidence that memorization inflates apparent accuracy at the cost of generalization.

Subjects:

Machine Learning (cs.LG)

Cite as: arXiv:2603.26797 [cs.LG]

(or arXiv:2603.26797v1 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2603.26797

arXiv-issued DOI via DataCite

Submission history

From: Dip Roy [view email] [v1] Thu, 26 Mar 2026 00:35:25 UTC (1,152 KB)