Multilingual Lm
Multilingual language models (MLMs) aim to build language understanding and generation capabilities across multiple languages, often leveraging transformer architectures. Current research focuses on improving MLM performance by optimizing data composition for training, investigating the relationship between model architecture and learnability across languages, and mitigating biases stemming from imbalanced training data and cross-lingual interference. These advancements are crucial for creating more robust and equitable natural language processing tools applicable to a wider range of languages and downstream tasks, impacting fields like machine translation, question answering, and cross-cultural communication.
Papers
External Language Model Integration for Factorized Neural Transducers
Michael Levit, Sarangarajan Parthasarathy, Cem Aksoylar, Mohammad Sadegh Rasooli, Shuangyu Chang
Impossible Distillation: from Low-Quality Model to High-Quality Dataset & Model for Summarization and Paraphrasing
Jaehun Jung, Peter West, Liwei Jiang, Faeze Brahman, Ximing Lu, Jillian Fisher, Taylor Sorensen, Yejin Choi