Abstract
As social-media platforms emerge and evolve faster than the regulations meant to oversee them, automated detoxification might serve as a timely tool for moderators to enforce safe discourse at scale. We here describe our submission to the PAN 2025 Multilingual Text Detoxification Challenge, which rewrites toxic single-sentence inputs into neutral paraphrases across 15 typologically diverse languages. Building on a 12B-parameter Gemma-3 multilingual transformer, we apply parameter-efficient LoRA SFT fine-tuning and prompting techniques like few-shot and Chain-of-Thought. Our multilingual training corpus combines 3,600 human-authored parallel pairs, 21,600 machine-translated synthetic pairs, and model-generated pairs filtered by Jaccard thresholds. At inference, inputs are enriched with three LaBSE-retrieved neighbors and explicit toxic-span annotations. Evaluated via Style Transfer Accuracy, LaBSE-based semantic preservation, and xCOMET fluency, our system ranks first on high-resource and low-resource languages. Ablations show +0.081 joint score increase from few-shot examples and +0.088 from basic CoT prompting. ANOVA analysis identifies language resource status as the strongest predictor of performance ($\eta^2$ = 0.667, p < 0.01).
Abstract (translated)
随着社交媒体平台的快速涌现和演变，监管措施往往跟不上步伐。在这种情况下，自动化去毒工具可以作为管理员在大规模范围内实施安全对话的重要手段。在这里，我们介绍了我们在PAN 2025多语言文本去毒挑战中的参赛作品，该作品将有毒的单句输入重写为15种类型多样化的语言中性短语。 我们的解决方案基于一个参数量达12B的Gemma-3多语言转换器，并应用了高效的LoRA SFT微调和如少样本学习（few-shot learning）及链式思维（Chain-of-Thought）等提示技术。我们使用的多语言训练数据集结合了3,600对由人类编写的平行文本、21,600对机器翻译的合成文本，以及通过杰卡德阈值过滤后生成的模型文本对。 在推理阶段，输入会使用LaBSE检索的三个邻居进行扩展，并加上明确的有毒片段注释。我们的系统在风格转换准确率（Style Transfer Accuracy）、基于LaBSE的语义保存能力和xCOMET流畅度方面进行了评估，在高资源和低资源语言中均排名第一。 消融研究表明，少样本示例可使联合得分增加0.081，而基本链式思维提示则使分数增加了0.088。方差分析（ANOVA）发现，语言资源状况是预测性能最强的因素（$\eta^2$ = 0.667, p < 0.01）。
URL
https://arxiv.org/abs/2510.01250
