Xiaoming Zhai

Interpreting and Steering LLMs with Mutual Information-based Explanations on Sparse Autoencoders

End-to-End On-Device Quantization-Aware Training for LLMs at Inference Cost

SketchMind: A Multi-Agent Cognitive Framework for Assessing Student-Drawn Scientific Sketches

Interpreting and Steering LLM Representations with Mutual Information-based Explanations on Sparse Autoencoders

An Efficient LLM Alignment Framework for Automated Radiology Impression Generation