Zhengyan Shi

Hi, weclome to my personal page. I am a PhD student supervised by Prof Aldo Lipani and Prof Emine Yilmaz at University College London, affiliated with Web Intelligence Group and SpaceTimeLab. Previously, I worked as an Intern of Technical Staff at Cohere in London and completed two internships as an Applied Scientist at Amazon in both London and Seattle offices.

Prior to pursuing my PhD, I obtained a Master's degree in Data Science (Statistics) with Distinction from University College London and a Bachelor's degree in Mathematics with First Class Honor from University of Liverpool and Xi'an Jiaotong-Liverpool University.

Central to my research is the ambition to leverage language models efficiently and robustly to solve general tasks. To that end, my existing work can be broadly categorized into the following directions:

Google Scholar  /  Twitter  /  Github  /  LinkedIn  /  Email

Profile photo

Research (Selected)

Paper preview

Instruction Tuning With Loss Over Instructions

Zhengyan Shi, Adam X. Yang, Bin Wu, Laurence Aitchison, Emine Yilmaz, Aldo Lipani

Advances in Neural Information Processing Systems (NeurIPS), 2024

We show that in certain scenarios, applying loss to instructions rather than outputs only, which we refer to as Instruction Modelling, could largely improve the performance of instruction tuning on both various NLP and open-ended generation benchmarks. Remarkably, in the most advantageous case, our approach boosts model performance on AlpacaEval 1.0 by over 100%.

Paper preview

Rethinking Semi-supervised Learning with Language Models

Zhengxiang Shi, Francesco Tonolini, Nikolaos Aletras, Emine Yilmaz, Gabriella Kazai, Yunlong Jiao

Association for Computational Linguistics (Findings of ACL), 2023

Shows Task-adaptive Pre-training (TAPT) as a simple yet effective method for semi-supervised learning (often SoTA performance). Highlights the effectiveness of TAPT even with only a few hundred unlabelled samples (in contrary to the common belief that continued pre-training requires a large amount of unlabelled data).

Paper preview

StepGame: A New Benchmark for Robust Multi-Hop Spatial Reasoning in Texts

Zhengxiang Shi, Qiang Zhang, Aldo Lipani

Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2022

Introduces StepGame, a new benchmark for testing multi-hop spatial reasoning in texts. This dataset challenges models to perform robust spatial reasoning across multiple steps, providing a valuable tool for advancing natural language understanding in complex spatial scenarios.

Teaching Activities

Academic Services

Program Committee: NeurIPS (2023, 2024), ICML (2024), ICLR (2025), AAAI (2023, 2024), COLM (2024), ACL ARR (Feb. 2023 - Jan. 2024), ACL (2023), EMNLP (2022, 2023), EACL (2023), COLING (2023, 2024), ECML/PKDD (2022), KDD (2023), SIGIR (2022, 2023, 2024), ECIR (2024), SDM (2024)