Reproduce “PersonaAgent: Retrieval-Augmented Transaction Anomaly Scoring with Refreshable Multi-Entity Persona Memory” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhengyu Fang, Liang Wang, Yingtong Dou, Yuzhong Chen, Shubham Jain, Junpeng Wang, Xiran Fan, Menghai Pan and 3 more
PersonaAgent adds shared behavioral memory to frozen anomaly detectors, using offline-designed records, weekly numerical updates, and deterministic lookups through a small adapter at scoring time. Across four deployed detectors, all 24 metric comparisons improved, with confidence intervals above zero. For Detector 2 in Period 2, BM1 rose 14.24% relative; this measures score quality, not fraud losses or changed authorizations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.