2025
Mitigating Hallucinated Translations in Large Language Models with Hallucination-focused Preference Optimization
NAACL 2025long
Machine Translation (MT) is undergoing a paradigm shift, with systems based on fine-tuned large language models (LLM) becoming increasingly competitive with traditional encoder-decoder models trained specifically for translation tasks. However, LLM-based systems are at a higher risk of generating ha…