![[2510.20102] Human-Centered LLM-Agent System for Detecting Anomalous Digital Asset Transactions](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2510.20102] Human-Centered LLM-Agent System for Detecting Anomalous Digital Asset Transactions
Summary
The paper presents HCLA, a human-centered multi-agent system designed for detecting anomalies in digital asset transactions, enhancing interpretability and decision transparency.
Why It Matters
As digital assets become more prevalent, the need for effective anomaly detection systems is critical for regulatory compliance and financial forensics. HCLA's approach emphasizes human involvement in the decision-making process, which is essential for building trust and accountability in high-stakes financial environments.
Key Takeaways
- HCLA integrates three roles for anomaly detection: Rule Abstraction, Evidence Scoring, and Expert-Style Justification.
- The system allows non-experts to engage in analytical processes using natural language.
- HCLA enhances interpretability and transparency in decision-making for digital asset transactions.
- The framework separates evidence scoring from justification, promoting accountability.
- Experiments demonstrate strong predictive accuracy and improved user interaction.
Computer Science > Artificial Intelligence arXiv:2510.20102 (cs) [Submitted on 23 Oct 2025 (v1), last revised 16 Feb 2026 (this version, v2)] Title:Human-Centered LLM-Agent System for Detecting Anomalous Digital Asset Transactions Authors:Gyuyeon Na, Minjung Park, Hyeonjeong Cha, Sangmi Chai View a PDF of the paper titled Human-Centered LLM-Agent System for Detecting Anomalous Digital Asset Transactions, by Gyuyeon Na and 3 other authors View PDF Abstract:We present HCLA, a human-centered multi-agent system for anomaly detection in digital-asset transactions. The system integrates three cognitively aligned roles: Rule Abstraction, Evidence Scoring, and Expert-Style Justification. These roles operate in a conversational workflow that enables non-experts to express analytical intent in natural language, inspect structured risk evidence, and obtain traceable, context-aware reasoning. Implemented with an open-source, web-based interface, HCLA translates user intent into explicit analytical rules, applies classical anomaly detectors to quantify evidential risk, and reconstructs expert-style justifications grounded in observable transactional signals. Experiments on a cryptocurrency anomaly dataset show that, while the underlying detector achieves strong predictive accuracy, HCLA substantially improves interpretability, interaction, and decision transparency. Importantly, HCLA is not designed to explain a black-box model in the conventional XAI sense. Instead, we reconstruct a t...
