Hang Yan

A new method to align AI responses with human preferences efficiently.

A new approach tackles the issue of dropped tokens and padding in machine learning models.

AnyGPT processes text, images, music, and speech for diverse applications.

This article explores how comment generation improves large language models’ coding abilities.

A high-quality dataset for training language models from English web content.

CHARM assesses how well LLMs reason in both global and Chinese contexts.

A new model that enhances code generation using multi-source data.

A new framework improves instruction data quality for language models.