The field of natural language processing is experiencing a significant shift towards improving support for low-resource languages. Recent developments have focused on leveraging multistage, multilingual, and domain-specific methods to enhance the performance of large language models in these languages. The use of continual pre-training and intermediate task transfer learning has shown promise in adapting models to extremely low-resource settings. Additionally, the importance of collecting and utilizing domain-specific data, even on a small scale, has been highlighted as a crucial factor in improving machine translation performance. Furthermore, the evaluation of large language models using language proficiency exams has emerged as a viable method for assessing their capabilities in low-resource languages. Overall, the field is moving towards a more inclusive and diverse approach to natural language processing, with a growing recognition of the need to support and develop models for low-resource languages. Noteworthy papers include:

• A study on leveraging multistage and multilingual methods for low-resource machine translation, which proposed two novel approaches for adapting large language models.
• A paper on the limitations of religious data in machine translation for Guinea-Bissau Creole, which highlighted the importance of collecting domain-specific data for low-resource languages.
• A case study on the efficacy of a large language model in computer education, which demonstrated its potential for assisting in network security education.