# SAFE-QAQ: End-to-End Slow-Thinking Audio-Text Fraud Detection via Reinforcement Learning

Source: [SAFE-QAQ: End-to-End Slow-Thinking Audio-Text Fraud Detection via Reinforcement Learning](<https://www.sota2.com/research/paper/695c96e353fdd5c361743413>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-07T05:00:04.582Z

Authors: Peidong Wang, Zhiming Ma, Xin Dai, Yongkang Liu, Shi Feng, Xiaocui Yang, Wenxing Hu, Zhihao Wang, Mingjun Pan, Li Yuan, Daling Wang

Published: 2026-01-04T06:09:07.000Z

## Abstract

Existing fraud detection methods predominantly rely on transcribed text, suffering from ASR errors and missing crucial acoustic cues like vocal tone and environmental context. This limits their effectiveness against complex deceptive strategies. To address these challenges, we first propose \\textbf{SAFE-QAQ}, an end-to-end comprehensive framework for audio-based slow-thinking fraud detection. First, the SAFE-QAQ framework eliminates the impact of transcription errors on detection performance. Secondly, we propose rule-based slow-thinking reward mechanisms that systematically guide the system to identify fraud-indicative patterns by accurately capturing fine-grained audio details, through hierarchical reasoning processes. Besides, our framework introduces a dynamic risk assessment framework during live calls, enabling early detection and prevention of fraud. Experiments on the TeleAntiFraud-Bench demonstrate that SAFE-QAQ achieves dramatic improvements over existing methods in multiple key dimensions, including accuracy, inference efficiency, and real-time processing capabilities. Currently deployed and analyzing over 70,000 calls daily, SAFE-QAQ effectively automates complex fraud detection, reducing human workload and financial losses. Code: https://anonymous.4open.science/r/SAFE-QAQ.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Telecom Fraud Analysis on TeleAntiFraud-Bench](<https://www.sota2.com/research/sota/telecom-fraud-analysis-on-teleantifraud-bench.md>): Telecom Fraud Analysis; TeleAntiFraud-Bench; Weighted F1 (Sce.): 84.64.

- [Scam Detection Benchmark Single-App Content](<https://www.sota2.com/research/sota/scam-detection-benchmark-single-app-content.md>): Scam Detection; Scam Detection Benchmark Single-App Content; Accuracy: 98.4.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2601.01392>)

[PDF](<https://arxiv.org/pdf/2601.01392.pdf>)
