Li Du

Explores the importance of tokenization methods in enhancing natural language processing models.

A concise overview of Byte-Pair Encoding and its role in NLP.

A new method improves how models perceive depth and spatial relationships in images.

This study examines how different data sources affect large language models.

AquilaMoE uses EfficientScale to optimize bilingual language model training with less data.

A new method improves language models by diagnosing knowledge deficiencies without labeled data.

This study uncovers how LLMs adapt their learning through attention patterns.