Yiming Tang

About Me

Hi! I'm Yiming, a third-year PhD student in National University of Singapore, very fortunate to be advised by Prof. Dianbo Liu. I lead the interpretability research group in Artificial Scientific Intelligence Lab. My research primarily aims to develop algorithms and theories toward a better understanding of intelligence for both scientific research and AI safety, currently centered around the identification of human-interpretable concepts encoded in LMMs, and shifting toward tracking representation drifts induced by post-training. During my undergraduate studies, I was very lucky to collaborate with Prof. Bin Dong, Prof. Shanghang Zhang, and Prof. Hao Dong.

News

• [2026.7] I have passed all of my qualifying examinations for my PhD program at NUS. 🎉
• [2026.7] I have recieved my certification for the Technical AI Safety Project from Bluedot. 🎉
• [2026.6] One paper accepted in Transactions on Machine Learning Research (SAEGA) . 🎉
• [2026.6] I have joined Alibaba Security for an internship on AIGC and AI Agent Safety. ✈️
• [2026.6] Our team is achknowledged and selected in the AI Safety Research Directory . 🎉
• [2026.3] One paper accepted (spotlight) in CVPR XAI4CV Workshop (Matryoshka Transcoder) . 🎉
• [2025.12] Our work, LanSE, has been selected in NUS Invision's SIMFONI programme . 🎉
• [2025.9] One paper finished and preprinted on arXiv (CXR-LanIC) . 🎉
• [2025.1] One paper accepted in International Conference on Machine Learning (SAN) . 🎉
• [2024.5] One paper finished and preprinted on arXiv (Demonstration Notebook) . 🎉
• [2024.3] One paper accepted in European Conference on Computer Vision (LLM Dataset Analyst) . 🎉
• [2023.10] One paper accepted in Journal of Machine Learning (PE Lens) . 🎉
• [2023.5] One paper finished and preprinted on arXiv (IntFF) . 🎉

Research Interests

My research interests focus on:

• Mechanistic Interpretability: Understanding the internal representation space of machine learning models, particularly vision-language models, by identifying specific concepts encoded within them.
• AI Safety: Assessing and benchmarking the safety and trustworthiness of AI systems and their generated contents, frequently utilizing interpretability modules and techniques.
• Representation Learning: Enhancing representation learning algorithms based on our understanding of the internal representation space of machine learning models.
• Reinforcement Learning: Studying how reinforcement learning algorithms changes the internal representation space of LLMs and LMMs.

Beyond these topics, I am also broadly interested in LMMs and AI Agents. See What We Work On.

Collaboration

I'm highly open for mutually beneficial collaborations with the central aim to advance AI Interpretability and AI safety via scientific research. Researchers working on aligned research topics are especially welcome. I can also offer research intern positions for undergraduate and master students. Feel free to contact me directly via email or WeChat if you are interested in our research.

Selected Works

A Unified Theory of Sparse Dictionary Learning in Mechanistic Interpretability: Piecewise Biconvexity and Spurious Minima

We develop the first theoretical framework for sparse dictionary learning in mechanistic interpretability.

Paper

How does My Model Fail? Automatic Identification and Interpretation of Physical Plausibility Failure Modes with Matryoshka Transcoders

We develop matryoshka transcoders and utilize it to identify physical plausibility failure modes of generative models, achieving SOTA performances in targeted feature discovery.

Paper

Human-like Content Analysis for Generative AI with Language-Grounded Sparse Encoders

We develop LanSE and decompose natural and medical images into interpretable visual patterns grounded in natural language, supporting fine-grained analysis on AI generated contents.

Paper

LLM as Dataset Analyst: Subpopulation Structure Discovery with Large Language Model

We are the first in the literature to prompt engineer LLMs to inspect datasets and analyze their subpopulation structures, paving the way for advanced dataset analysis with LLMs.

Paper

Prompt Engineering Through the Lens of Optimal Control

We are one of the first approaches to develop a theoretical framework to understand various prompt engineering methods through the lens of rigorous optimal control theory.

Paper

Introducing Brainiac Buddy, your AI-powered teaching assistant

It was a real honor to participate in the early development of Brainiac Buddy, an AI-powered teaching assistant with real-world applications in Peking University developed by Bin Dong's team.

Project

Bridging Mechanistic Interpretability and Prompt Engineering with Gradient Ascent for Interpretable Persona Control

We are the first in the literature to apply gradient ascent on MechInterp features to support prompt engineering for persona control.

Paper

When Language Overwrites Vision: Over-Alignment and Geometric Debiasing in Vision-Language Models

We identify, analyze, and mitigate the existence of textual biases in visual tokens in vision-language models.

Paper

The Integrated Forward-Forward Algorithm: Integrating Forward-Forward and Shallow Backpropagation with Local Losses

We follow Hinton's Forward-Forward algorithm and make it applicable for deeper layers via the introduction of local losses.

Paper

Publications and Preprints

For a complete list of publications, please visit my Google Scholar profile.

Education

National University of Singapore

Doctor of Philosophy, College of Design and Engineering

Peking University

Bachelor of Science, School of Mathematical Sciences

Talks

Experiences

Research Intern

Alibaba Security, Alibaba Group Holding Limited, Hangzhou

June 2026 - Present

Supervisor: Hang Zhou

Research Associate

Artificial Scientific Intelligence Lab, National University of Singapore, Singapore

August 2024 - Present

Supervisor: Dianbo Liu

Research Intern

Human Machine Intelligence Lab, Peking University, Beijing

2023 - 2024

Supervisor: Shanghang Zhang

Research Intern

Beijing International Center for Mathematical Research, Peking University, Beijing

2023 - 2024

Supervisor: Bin Dong

Research Intern

Hyperplane Lab, Peking University, Beijing

2021 - 2022

Supervisor: Hao Dong

Teaching Researcher

Aijianzi, Beijing Aijianzi Education Technology, Beijing

2019 - 2021

Supervisor: Xiaojun Hu

Mentorship

I have had the privilege of mentoring several students working on research topics in Mechanistic Interpretability, Neuroscience, and AI for Biomedical Sciences.

Harshvardhan Saini

Indian Institute of Technology, Dhanbad, India

Zheng Lin

Hong Kong University of Science and Technology, Hong Kong, China

Zhaoqian Yao

Chinese University of Hong Kong, Hong Kong, China

Qinglin Qi

Lund University, Lund, Sweden

Lingheng Du

Peking University, Beijing, China

Yang Tuấn Anh

National University of Singapore, Singapore

Meet Desmond

An honor and a pleasure to introduce to you here my most adorable cat, Desmond, the docent of my research and the guardian of this website. Have a chat and he will answer your questions in my absence.

Pat Desmond Have a Chat!

Desmond is Tired! You Have Reached Maximum Patting Times per Visit. Come Back Later or Subscribe to Desmond Pro for Unlimited Pats (>w<).

Wellcome, visitor! Master is out, let us talk. 🐱

Send