Guanyu Yao | University of California, Santa Barbara

About Me

I am a second-year Ph.D. student at University of California, Santa Barbara, advised by Prof. Shiyu Chang. I received my B.Eng. degree in Artificial Intelligence from Beijing Normal University in 2020. My research interests lie in Multi-Modal LLMs and LLM Evaluation, I am also interested in LLM training and inference platforms.

Research Interests

Multi-Modal LLMs: multimodal pretraining, multimodal model architectures.
LLM Evaluation: evaluation metrics, evaluation datasets, evaluation platforms.

News

[2024-09-01] I joined UCSB as a Ph.D. student, advised by Prof. Shiyu Chang.
[2024-06-30] I graduated from Beijing Normal University with a B.Eng. degree in Artificial Intelligence.
[2023-03-01] I joined HKUST as an undergraduate research intern, advised by Prof. Tong Zhang.

Publications

An Improved Autoregressive Evaluation Paradigm for Large Language Models

Jipeng Zhang, Rui Pan, Yuzheng Hu, KaShun SHUM, Guanyu Yao, Xiang Liu, Renjie Pi, Hanze Dong, Shizhe Diao, Yong Lin, Han Zhao, Tong Zhang

PDF TIST 2025

Rethinking the Text-Vision Reasoning Imbalance in MLLMs through the Lens of Training Recipes

Guanyu Yao, Qiucheng Wu, Yang Zhang, Zhaowen Wang, Handong Zhao, Shiyu Chang

PDF arXiv 2025

Projects

LMFLow Core Contributor of an open source post-training platform(Github 8k+ 🌟) for large language models.

Teaching Experience

Teaching Assistant, CS 24: Problem Solving with Computers 2, UCSB, Fall 2025
Teaching Assistant, CS 190I: Generative AI, UCSB, Spring 2025
Teaching Assistant, CS 24: Problem Solving with Computers 2, UCSB, Winter 2025
Grader, CS 190I: Introduction to Deep Learning, UCSB, Fall 2024