可逆 LLM 规则叠加层：为冻结的反欺诈模型修补盲区
Patching the Blind Spots: Reversible LLM Rule Overlay for Frozen Fraud Detectors
AI 导读
一项 arXiv 研究提出用 LLM 推理修补已部署的梯度提升树反欺诈模型：LLM 读取模型树摘要与 SHAP 归因，生成规则并校准为生产阈值处的可逆 log-lift 叠加层，不改原模型。
正文 · 原文
Abstract:Gradient-boosted trees dominate industrial tabular fraud classification, but even a well-trained model has blind spots, and retraining a deployed model is not always safe. We bring LLM reasoning to patching such a model without modifying it: LLMs read its tree digest and SHAP attributions, reason about the fraud it misses, and propose rules anchored on features it underweights, screened and calibrated into a reversible log-lift overlay at the production cut-off. Single-LLM benefit is high-variance and inverts between datasets, so we pool three diverse backbones by logit averaging, trading peak gain for a bounded worst case. On an experimental detector trained on sampled, selection-biased real-world e-commerce data, tuned continued training regresses, whereas the overlay gains on average +2.26% recall at equal precision ($\Delta$R@P) and +1.60% precision at equal recall ($\Delta$P@R). Replayed offline on a legacy production detector, the overlay, calibrated on one month and frozen, stays more precise at equal recall for seven months (+5.64% $\Delta$P@R). On seven bias-shift variants of a public benchmark it raises $\Delta$P@R on all and keeps $\Delta$R@P near neutral. LLM reasoning over a model's diagnostics can improve a deployed tabular classifier through readable predicates naming the fraud they cover.
| Comments: | 12 pages, 1 figure, 8 tables |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.08714 [cs.LG] |
| (or arXiv:2611.08714v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.08714 arXiv-issued DOI via DataCite |
Submission history
From: Bo Qu [view email]
[v1]
Thu, 5 Nov 2026 21:12:11 UTC (35 KB)
[v2]
Wed, 11 Nov 2026 04:53:07 UTC (36 KB)
来源：arXiv cs.AI · arxiv.org