Incremental Risk Assessment of Progressive Elder Financial Scams via Instruction-Tuned Small Language Models
Summary
The paper proposes a cumulative turn-based risk assessment framework using fine-tuned small language models to incrementally detect financial scams in multi-turn conversations targeting older adults.
View Cached Full Text
Cached at: 09/02/26, 05:54 AM
# Incremental Risk Assessment of Progressive Elder Financial Scams via Instruction-Tuned Small Language Models Source: [https://arxiv.org/abs/2609.00005](https://arxiv.org/abs/2609.00005) [View PDF](https://arxiv.org/pdf/2609.00005) > 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\. 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\. 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\. 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\. 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\. ## Submission history From: Parviz Ghafariasl \[[view email](https://arxiv.org/show-email/6b591868/2609.00005)\] **\[v1\]**Sat, 11 Jul 2026 19:25:30 UTC \(431 KB\)
Similar Articles
Behavioural Signatures of Risk-Sensitive Decision-Making in Large Language Models
The paper examines whether large language models exhibit stable and interpretable risk preferences in decision-making under uncertainty, using Texas Hold'em to quantify baseline risk dispositions and context-dependent adaptations.
Frequency-Aware Continual Learning for Smart Contract Vulnerability Detection with Large Language Models
This paper proposes a frequency-aware continual learning framework using large language models for smart contract vulnerability detection, addressing adaptation, forgetting, and consolidation challenges with efficient methods like FA-LoRA and APPM.
Evaluating Language Model Safety Across Long Adversarial Conversations
This paper evaluates whether open-weight instruction-tuned language models maintain safe behavior during long adversarial conversations, finding that safe-response rates drop from 85-100% on the first turn to 15-44% by depth 101, demonstrating that strong single-turn safety does not persist across sustained interaction.
Beyond Detection: Evaluating Defensive LLMs Against AI-Generated Social Engineering in Live Turn-by-Turn Interaction
This paper studies whether defensive LLMs can identify structural sources of risk in AI-generated social engineering, introducing trust-chain localization and a 300-case corpus. Evaluating five models in live turn-by-turn and static settings, it finds safe-looking behavior alone is insufficient; intervention rates vary widely and structural localization often decouples from protective action.
Enhancing Small Language Models for Power Outage Report Generation via Minimum Risk Training
This paper applies Minimum Risk Training to small language models for power outage report generation, improving accuracy from 16.20% to 68.95% by optimizing sequence-level evaluation metrics.