Haoyu He (何灏宇)

prof_pic.jpg

Tübingen AI Center

Tübingen, Germany

I am a third-year PhD student in the Autonomous Vision Group (AVG) supervised by Prof. Andreas Geiger at Univeristy of Tübingen and Tübingen AI Center.

My research interest originates with language modeling, driven by a deep curiosity in understanding how machines can learn to interact with human. My current research focuses on pushing the shift from traditional Transformer-based models under the ‘linearization assumption (everything in the inputs is flattened)’ to models that can use ubiquitous hierarchies effectively.

I am also interested in multimodal modeling, and I draw a lot of sustenance from our exceptional CV group. I feel that though current LLMs are powerful at retrieving knowledge, they are far from an autonomous agent, partially because the lack of understanding laws (e.g., Physics) of how the world operates, which other modalities (e.g., images) can be complementary.

My vision naive dream for AI is to free humans to fully create and experience the way of life. Naturally, there are many obstacles along the way…

My CV is here

news

Oct 31, 2024 Our paper “NN4SysBench: Characterizing Neural Network Verification for Computer Systems” is accepted to NeurIPS 2024! Code and pdfs will be released soon.
Jul 10, 2024 Our paper “HDT: Hierarchical Document Transformer” is accepted to COLM 2024!

selected publications

  1. HDT: Hierarchical Document Transformer
    Haoyu He, Markus Flicke, Jan Buchmann, and 2 more authors
    In First Conference on Language Modeling, 2024
  2. Characterizing Neural Network Verification for Systems with NN4SYSBench
    Haoyu He, Tianhao Wei, Huan Zhang, and 2 more authors
    1st Workshop on Formal Verification of Machine Learning, 2022
  3. Distiller: A Systematic Study of Model Distillation Methods in Natural Language Processing
    Haoyu He, Xingjian Shi, Jonas Mueller, and 3 more authors
    In Proceedings of the Second Workshop on Simple and Efficient Natural Language Processing, Nov 2021