Jackie CK Cheung

Learning Task-Agnostic Representations through Multi-Teacher Distillation

Partial Perspectives: How LLMs Handle Logically Inconsistent Knowledge in Reasoning Tasks

Identifying and Analyzing Task-Encoding Tokens in Large Language Models

How to distill task-agnostic representations from many teachers?

When is an Embedding Model More Promising than Another?

Ensemble Distillation for Unsupervised Constituency Parsing

Do LLMs Build World Representations? Probing Through the Lens of State Abstraction

Successor Features for Efficient Multi-Subject Controlled Text Generation