Incremental Risk Assessment of Progressive Elder Financial Scams via Instruction-Tuned Small Language Models

基于指令微调小型语言模型的渐进式老年金融诈骗风险增量评估

Abstract: Financial scams targeting older adults increasingly occur through text and voice channels such as email, SMS, and phone calls, unfolding over multiple conversational turns that begin with impersonation or casual contact, escalate through trust building and urgency, and culminate in requests for sensitive information or financial transfers.

摘要： 针对老年人的金融诈骗越来越多地通过电子邮件、短信和电话等文本及语音渠道发生。这些诈骗通常通过多轮对话展开，从冒充身份或随意接触开始，通过建立信任和制造紧迫感逐步升级，最终以索取敏感信息或要求转账告终。

Because risk signals emerge incrementally across turns, effective detection requires models that continuously update risk estimates under resource-constrained deployment settings. We propose a cumulative turn-based risk assessment framework that incrementally aggregates conversational turns and re-estimates risk at each step, enabling dynamic scam monitoring across progressively evolving conversations.

由于风险信号在对话过程中是逐步显现的，有效的检测需要模型能够在资源受限的部署环境下持续更新风险评估。我们提出了一种基于累积轮次的风险评估框架，该框架通过增量聚合对话轮次并在每一步重新评估风险，从而实现对不断演变的对话进行动态诈骗监测。

A multi-turn dialogue dataset is constructed to cover investment, charity, and tech support scam scenarios, with each dialogue containing two to eight turns and annotated at every cumulative stage with a qualitative risk level, a continuous risk score, an explanatory rationale, and a safety recommendation.

我们构建了一个涵盖投资、慈善和技术支持诈骗场景的多轮对话数据集。每段对话包含两到八轮，并在每个累积阶段都标注了定性风险等级、连续风险评分、解释性理由以及安全建议。

Four small language models (Phi-4, LLaMA-3.2, DeepSeek-R1, and Qwen3) are fine-tuned and evaluated under a unified training framework. Fine-tuned small models capture fraud-related linguistic cues and cross-turn escalation patterns while maintaining compact architectures suitable for mobile and resource-constrained deployment settings.

我们在统一的训练框架下对四种小型语言模型（Phi-4、LLaMA-3.2、DeepSeek-R1 和 Qwen3）进行了微调和评估。微调后的小型模型能够捕捉与欺诈相关的语言线索和跨轮次升级模式，同时保持紧凑的架构，非常适合移动端和资源受限的部署环境。

Among the evaluated models, Phi-4 and LLaMA-3.2 achieve stronger turn-aware risk estimation performance relative to their parameter scale. These results suggest that structured cumulative modeling can support incremental scam risk assessment in deployment-oriented settings while highlighting the potential of compact language models for privacy-aware and on-device fraud protection.

在评估的模型中，Phi-4 和 LLaMA-3.2 相较于其参数规模，展现出了更强的轮次感知风险评估性能。这些结果表明，结构化的累积建模可以支持面向部署环境的增量诈骗风险评估，同时也凸显了紧凑型语言模型在注重隐私保护和端侧欺诈防护方面的潜力。