Saved in:
Bibliographic Details
Main Authors: Chua, Gabriel, Tan, Leanne, Ge, Ziyu, Lee, Roy Ka-Wei
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.05980
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910008092917760
author Chua, Gabriel
Tan, Leanne
Ge, Ziyu
Lee, Roy Ka-Wei
author_facet Chua, Gabriel
Tan, Leanne
Ge, Ziyu
Lee, Roy Ka-Wei
contents Large language models (LLMs) often fail to maintain safety in low-resource language varieties, such as code-mixed vernaculars and regional dialects. We introduce RabakBench, a multilingual safety benchmark and scalable pipeline localized to Singapore's unique linguistic landscape, covering Singlish, Chinese, Malay, and Tamil. We construct the benchmark through a three-stage pipeline: (1) Generate: augmenting real-world unsafe web content via LLM-driven red teaming; (2) Label: applying semi-automated multi-label annotation using majority-voted LLM labelers; and (3) Translate: performing high-fidelity, toxicity-preserving translation. The resulting dataset contains over 5,000 examples across six fine-grained safety categories. Despite using LLMs for scalability, our framework maintains rigorous human oversight, achieving 0.70-0.80 inter-annotator agreement. Evaluations of 13 state-of-the-art guardrails reveal significant performance degradation, underscoring the need for localized evaluation. RabakBench provides a reproducible framework for building safety benchmarks in underserved communities.
format Preprint
id arxiv_https___arxiv_org_abs_2507_05980
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Lost in Localization: Building RabakBench with Human-in-the-Loop Validation to Measure Multilingual Safety Gaps
Chua, Gabriel
Tan, Leanne
Ge, Ziyu
Lee, Roy Ka-Wei
Computation and Language
Machine Learning
Large language models (LLMs) often fail to maintain safety in low-resource language varieties, such as code-mixed vernaculars and regional dialects. We introduce RabakBench, a multilingual safety benchmark and scalable pipeline localized to Singapore's unique linguistic landscape, covering Singlish, Chinese, Malay, and Tamil. We construct the benchmark through a three-stage pipeline: (1) Generate: augmenting real-world unsafe web content via LLM-driven red teaming; (2) Label: applying semi-automated multi-label annotation using majority-voted LLM labelers; and (3) Translate: performing high-fidelity, toxicity-preserving translation. The resulting dataset contains over 5,000 examples across six fine-grained safety categories. Despite using LLMs for scalability, our framework maintains rigorous human oversight, achieving 0.70-0.80 inter-annotator agreement. Evaluations of 13 state-of-the-art guardrails reveal significant performance degradation, underscoring the need for localized evaluation. RabakBench provides a reproducible framework for building safety benchmarks in underserved communities.
title Lost in Localization: Building RabakBench with Human-in-the-Loop Validation to Measure Multilingual Safety Gaps
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2507.05980