FRAUDSkill: Structured Frozen-Weight Skill Optimization for Audio Anti-Fraud Detection
FRAUDSkill adapts frozen audio-language models for telecom fraud detection via external skill programs, reaching 73.50% Macro-F1.
FRAUDSkill is a structured frozen-weight adaptation framework for audio anti-fraud detection that optimizes external skill programs, route-specific policies, and decision rules without modifying the underlying audio-language model. On the TeleAntiFraud benchmark it achieves 73.50% Macro-F1, a 31.96% improvement over the shared frozen-model baseline, while reducing invalid outputs to 1.94%. Source code is publicly available.
- External skill optimization leaves the base audio-language model unchanged
- 73.50% Macro-F1 on TeleAntiFraud, +31.96% over frozen baseline
- Reduces invalid outputs to 1.94% via structured output control
- Uses validation-guided multi-path inference for protocol compliance
Full article172 words · extracted from huggingface.co · click to collapse
Large audio-language models have shown promise for anti-fraud detection by directly processing speech and reasoning over fraud-related evidence. Their deployment, however, requires predictions to follow a predefined label space and a structured decision protocol consisting of service-scenario identification, fraud detection, and conditional fraud-type classification. Existing fine-tuning and prompt-based approaches typically encode task knowledge, constraints, and decision rules into model parameters or manually maintained prompts, making them difficult to adapt as fraud patterns and labeling policies evolve. To this end, we propose FRAUDSkill, a structured frozen-weight adaptation framework that leaves the underlying audio-language model unchanged while optimizing an external layer of skill programs, route-specific policies, and decision rules. We further combine structured output control with validation-guided multi-path inference to ensure protocol-compliant predictions. On the TeleAntiFraud benchmark, FRAUDSkill achieves 73.50% Macro-F1, outperforming the shared frozen-model baseline by 31.96% while reducing invalid outputs to 1.94%. Extensive experiments demonstrate that external skill optimization provides an effective and adaptable solution for structured audio anti-fraud detection without modifying the underlying model. The source code is available at https://anonymous.4open.science/r/FRAUDSKILL-114514.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.18766