System Report for CCL25-Eval Task 10: SRAG-MAV for Fine-Grained Chinese Hate Speech Recognition

TL;DR

This paper introduces SRAG-MAV, a novel framework combining task reformulation, retrieval-augmented generation, and multi-round voting to enhance fine-grained Chinese hate speech recognition, achieving state-of-the-art results on the STATE ToxiCN dataset.

Contribution

The paper proposes a new SRAG-MAV framework that integrates multiple techniques to improve Chinese hate speech recognition performance.

Findings

Achieved a Hard Score of 26.66 and Soft Score of 48.35 on the STATE ToxiCN dataset.

Outperformed baselines including GPT-4o and fine-tuned Qwen2.5-7B.

Demonstrated significant performance improvements with the proposed methods.

Abstract

This paper presents our system for CCL25-Eval Task 10, addressing Fine-Grained Chinese Hate Speech Recognition (FGCHSR). We propose a novel SRAG-MAV framework that synergistically integrates task reformulation(TR), Self-Retrieval-Augmented Generation (SRAG), and Multi-Round Accumulative Voting (MAV). Our method reformulates the quadruplet extraction task into triplet extraction, uses dynamic retrieval from the training set to create contextual prompts, and applies multi-round inference with voting to improve output stability and performance. Our system, based on the Qwen2.5-7B model, achieves a Hard Score of 26.66, a Soft Score of 48.35, and an Average Score of 37.505 on the STATE ToxiCN dataset, significantly outperforming baselines such as GPT-4o (Average Score 15.63) and fine-tuned Qwen2.5-7B (Average Score 35.365). The code is available at https://github.com/king-wang123/CCL25-SRAG-MAV.