{"ID":2892123,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2507.15339","arxiv_id":"2507.15339","title":"LionGuard 2: Building Lightweight, Data-Efficient \u0026 Localised Multilingual Content Moderators","abstract":"Modern moderation systems increasingly support multiple languages, but often fail to address localisation and low-resource variants - creating safety gaps in real-world deployments. Small models offer a potential alternative to large LLMs, yet still demand considerable data and compute. We present LionGuard 2, a lightweight, multilingual moderation classifier tailored to the Singapore context, supporting English, Chinese, Malay, and partial Tamil. Built on pre-trained OpenAI embeddings and a multi-head ordinal classifier, LionGuard 2 outperforms several commercial and open-source systems across 17 benchmarks, including both Singapore-specific and public English datasets. The system is actively deployed within the Singapore Government, demonstrating practical efficacy at scale. Our findings show that high-quality local data and robust multilingual embeddings can achieve strong moderation performance, without fine-tuning large models. We release our model weights and part of our training data to support future work on LLM safety.","short_abstract":"Modern moderation systems increasingly support multiple languages, but often fail to address localisation and low-resource variants - creating safety gaps in real-world deployments. Small models offer a potential alternative to large LLMs, yet still demand considerable data and compute. We present LionGuard 2, a lightw...","url_abs":"https://arxiv.org/abs/2507.15339","url_pdf":"https://arxiv.org/pdf/2507.15339v2","authors":"[\"Leanne Tan\",\"Gabriel Chua\",\"Ziyu Ge\",\"Roy Ka-Wei Lee\"]","published":"2025-07-21T07:50:48Z","proceeding":"cs.CL","tasks":"[\"cs.CL\",\"cs.LG\"]","methods":"[\"Large Language Model\"]","has_code":false}