Yimeng Gu

I am a final-year Computer Science PhD student at Queen Mary University of London, where I am advised by Prof. Gareth Tyson.

Prior to joining Queen Mary, I obtained my B.E. from Beihang University and my two M.S. from Carnegie Mellon University and The Hong Kong University of Science and Technology.

I work on natural language processing with an application in misinformation detection. I was a research intern in Autodesk AI lab during summer 2023.

I am actively seeking industry opportunities.

Email  /  CV  /  Google Scholar

Apr 2025: One paper accepted to IJCNN 2025.
Feb 2025: One paper accepted to Pattern Recognition.
Jan 2024: One paper accepted to ACM WebSci 2024.
Nov 2023: One paper accepted to ICWSM 2024.
Apr 2023: I will be interning at Autodesk Research this summer.
Mar 2022: I ranked 16/69 on sub-task A of SemEval 2022 Task 5: Multimedia Automatic Misogyny Identification.

I'm interested in both multimodal learning and its applications, and the broad area of natural language processing.

TBD

We propose ConDA-TTT to learn the domain-invariant features for out-of-context detection.

We propose GatedVAE (Gated Variational AutoEncoder), which enables VAE with the gating mechanism, in order to dynamically let pass the noisy modality.

This paper studies the growing use of professional editor-curation for user-generated comments. We further propose a set of models that can automatically identify good candidate editor-picks.

We propose a Multi-modal Multi-task Variational AutoEncoder (MMVAE) to learn an effective co-representation of visual and textual features of memes in the latent space, and determine if the meme contains misogynous information and identify its fine-grained categories.

We first create a large dataset known as CMUmine™ and then demonstrate that, using NLP and ML techniques the claim construction process in patent applications can be automated.

We build AutodEncoder + latentGAN to learn the probablistic distributions of the neighbouring parts of a given part in the assembly. We evaluate the model performance both quantitively (IoU) and qualitatively. Our approach is able to predict the neighboring parts for the part query from unseen datasets.

ACM 16th Web Science Conference [May 2024]: Detecting Multimodal Fake News with Gated Variational AutoEncoder
British Machine Vision Association [Apr 2024]: Learning Domain-Invariant Feature for Out-of-context News Detection
Autodesk Research Connections [Aug 2023]: Identifying Mechanisms in Fusion360 Assemblies

ECS765P Big Data Processing - Spring 2022 (TA)

In my spare time, I like playing tennis, badminton, ping-pong, basketball and working out. I like watching almost all kinds of sport games.

I'm also a museum lover, especially for natural history museums and museums related to humanity culture. Some cool museums I have been to: the Qsingdao Beer Museum, the BMW Vehicle Museum, the Mercedes-Benz Museum.

Last update on Apr 1st, 2025. Template credits to Jon Barron.