Biography

Hi, I’m Keyu, a master student in Machine Learning at University of Tuebingen. Currently, I am a research assistant in Max Planck Institute for Intelligent Systems, advised by Dr. Shiwei Liu, and also have the privilege of working with Dr. Weiyang Liu and Dr. Jonas Geiping. I earned my B.Eng. in Artificial Intelligence from Southeast University where I was advised by Prof. Guilin Qi, and had a wonderful time at SEU-131AIClub, contributing to its founding and flourishing.

I am passionate about the empirical wonders from how machines learn – on what data, at what scale, stepping by what rule, and how early choices set the ceiling on foundamental capabilities like generalization and lifelong learning. My current research centers on scalable and efficient foundation models, with a focus on understanding their data, architectures and optimisation, and their effect on training dynamics.

I am seeking PhD opportunities in 27 fall/winter and internship during 26 winter/27 spring. Here are my Research-CV (last updated on 30. Aug). Please feel free to reach out to me via email at keyu.wang@student.uni-tuebingen.de.

Education

2024.10 - Present, M.Sc. in Machine Learning, University of Tuebingen, Tuebingen, Germany
2020.09 - 2024.06, B.Eng. in Artificial Intelligence, Southeast University, Nanjing, China

Working Experience

2025.12 - 2027.01, Research Assistant, Max Planck Institute for Intelligent Systems, Tuebingen, Germany
2023.07 - 2024.06, Intern, BSH – Bosch-Siemens Home Appliance Group, Nanjing, China

Recent Publications

Refer to my Google Scholar for a complete list. (* Equal contribution)

I. Learning: Training Dynamics, Data, and Optimization

DepthBench: Measuring How Residual Connections Enable More Computational Depth

Keyu Wang*, Yangyi Huang*, Jiale Kang, David González-Martínez, Weiyang Liu and Shiwei Liu
arXiv 2026 [PDF]

One LR Doesn’t Fit All: Heavy-Tail Guided Layerwise Learning Rates for LLMs

Di He, Songjun Tu, Keyu Wang, Lu Yin and Shiwei Liu
ICML 2026 [PDF]

II. Inference: Evaluation, Efficiency, and Reasoning Reliability

When Fewer Layers Break More Chains: Layer Pruning Harms Test-Time Scaling in LLMs

Keyu Wang*, Tian Lyu*, Guinan Su, Lu Yin, Marco Canini, Jonas Geiping, Shiwei Liu
COLM 2026 [PDF]

Academic Services

Conference Reviewer: COLM 2026, CPAL 2026
Workshop Reviewer: On-device Intelligence @ NeurIPS 2026

Misc

My Chinese name is 汪可予. I come from Tongling, Anhui — a small city by the Yangtze River.
I like all kinds of sports, such as table tennis, basketball, football, badminton, marathons, jogging, hiking and climbing.
I a big fan of animations. I enjoy various chess, board and card games.