L4NLP

AI & ML interests

L4NLP (Long-context for NLP applications) lab is part of the HKUNLP group. We focus on (1) developing new approaches that extend the context window of large language models while maintaining the reasoning ability of LLMs on the pertaining context length, (2) Instituting standardized evaluation methods for Long-context language models and (3) collecting long SFT data for training long-context models.

Recent Activity

Sansa  authored a paper 2 months ago

L-Eval: Instituting Standardized Evaluation for Long Context Language Models

Sansa  authored a paper 2 months ago

BBA: Bi-Modal Behavioral Alignment for Reasoning with Large Vision-Language Models

Sansa  authored a paper 2 months ago

Training-Free Long-Context Scaling of Large Language Models

View all activity

Team members 4