Tag
文章探讨了大型AI模型在小语种和地方方言上的表现不佳问题：训练数据不足导致翻译生硬怪异，并可能把用户推向主流语言。文章提出AI公司是否应承担支持小语种的责任，即使缺乏商业回报。
The article proposes that type diversity in datasets explains why Transformers struggle with structural compositional generalisation, and validates this by varying diversity in COGS and SLOG datasets.
This research paper examines how large language models contribute to the decline of linguistic diversity, highlighting potential cultural and societal risks.
This paper presents Structure-Guided Entity Resolution (SGER), a framework that fine-tunes LLMs through curriculum learning for robust person name matching in linguistically diverse contexts, achieving 99.02% accuracy on Indian identity data and deployed at Dream11.