Towards Reliable Machine Translation: Scaling LLMs for Critical Error Detection and Safety
This work addresses the need for safer and more trustworthy multilingual systems, particularly in high-stakes or underrepresented contexts, by improving error detection to reduce risks like disinformation and miscommunication, though it appears incremental as it builds on existing LLM methods.
The paper tackles the problem of critical meaning errors in machine translation, such as factual distortions and biased translations, by exploring instruction-tuned Large Language Models (LLMs) for detection, finding that model scaling and adaptation strategies outperform encoder-only baselines like XLM-R and ModernBERT.
Machine Translation (MT) plays a pivotal role in cross-lingual information access, public policy communication, and equitable knowledge dissemination. However, critical meaning errors, such as factual distortions, intent reversals, or biased translations, can undermine the reliability, fairness, and safety of multilingual systems. In this work, we explore the capacity of instruction-tuned Large Language Models (LLMs) to detect such critical errors, evaluating models across a range of parameters using the publicly accessible data sets. Our findings show that model scaling and adaptation strategies (zero-shot, few-shot, fine-tuning) yield consistent improvements, outperforming encoder-only baselines like XLM-R and ModernBERT. We argue that improving critical error detection in MT contributes to safer, more trustworthy, and socially accountable information systems by reducing the risk of disinformation, miscommunication, and linguistic harm, especially in high-stakes or underrepresented contexts. This work positions error detection not merely as a technical challenge, but as a necessary safeguard in the pursuit of just and responsible multilingual AI. The code will be made available at GitHub.