作者位次 #3/7ACL 2026Poster· Natural Language Processing
TOWER +: Bridging Generality and Translation Specialization in Multilingual LLM s
Ricardo Rei · Nuno M Guerreiro · José Pombal · João Alves · Amin Farajian · Pedro Teixeirinha · Andre Martins
摘要
Fine-tuning pretrained LLMs has proven effective for reaching state-of-the-art performance on specific tasks like machine translation. However, this process often implies sacrificing general-purpose capabilities, such as conversational reasoning and instruction-following, hampering the usefulness...