Hi 👋, I'm Lang /læŋ/

I am currently a first-year PhD student at MBZUAI, a great place for research, and an algorithm intern at ByteDance. I’m fortunate to be supervised by Dr. Xiuying Chen, an outstanding rising star and a truly supportive mentor. I am also co-supervised by the excellent Prof. Preslav Nakov.

Previously, I received my bachelor’s degree in Computer Science and Technology from Huazhong University of Science and Technology (HUST) in 2025.

💡 Interests

Primary

🔬 Mechanistic Interpretability

To know the mechanistic reasons why foundation models can do some things and cannot do others, and how to make them do what we want by utilizing their nature, with a particular focus on representation geometry and representation learning. This line of research is valuable in two ways:

Making foundation models more interpretable, controllable, and trustworthy;
Offering new perspectives and application logic for humans to explore AI.

Recently, I have been particularly into the geometrical features of latent spaces and ways to escape from the Linear Representation Hypothesis.

Secondary

🛡️ Trustworthy AI

Investigating the safety, fairness, and reliability of LLMs, including defending against jailbreak attacks, detecting and mitigating social biases within model representations, and improving the robustness of machine-generated text detection.

📝 Selected Publications

View Full Publication List on Google Scholar

🧑‍🔬 Mechanistic Interpretability

The Cylindrical Representation Hypothesis for Language Model Steering

Lang Gao, Jinghui Zhang, Wei Liu, Fengxian Ji, Chenxi Wang, Zirui Song, Akash Ghosh, Youssef Mohamed, Preslav Nakov, and Xiuying Chen

“By relaxing the orthogonality assumption in the Linear Representation Hypothesis, we introduce a cylindrical structure that explains why steering directions often interact and lead to unstable behaviors.”

Paper Code Dataset Talk@Cohere.ai

When Personalization Tricks Detectors: The Feature-Inversion Trap in Machine-Generated Text Detection

Lang Gao, Xuhui Li, Chenxi Wang, Mingzhe Li, Wei Liu, Zirui Song, Jinghui Zhang, Rui Yan, Preslav Nakov, and Xiuying Chen

“What if I told you your AI detector can still achieve a high AUC on random tokens? The first work to reveal the weak transferability of machine-generated text detectors on personalized content, along with its mechanistic interpretation.”

Paper Data & Code Talk@ 机器之心

Evaluate Bias without Manual Test Sets: A Concept Representation Perspective for LLMs

Lang Gao, Kaiyang Wan, Wei Liu, Chenxi Wang, Zirui Song, Zixiang Xu, Yanbo Wang, Veselin Stoyanov, and Xiuying Chen

“BiasLens is a new interpretable method that directly examines concept representations inside LLMs to detect hidden biases, without relying on any human-labeled data.”

Paper Code

Shaping the Safety Boundaries: Understanding and Defending Against Jailbreaks in Large Language Models

Lang Gao, Jiahui Geng, Xiangliang Zhang, Preslav Nakov, and Xiuying Chen

“We interpret the common mechanisms of diverse LLM jailbreak attacks in the activation space and propose an efficient defense method.”

Paper MBZUAI News

A Fano-Style Accuracy Upper Bound for LLM Single-Pass Reasoning in Multi-Hop QA

Kaiyang Wan, Lang Gao, Honglin Mu, Preslav Nakov, Yuxia Wang, Xiuying Chen

“An LLM’s accuracy suffers a severe breakdown once the required information exceeds its internal capacity in complex multi-hop reasoning scenarios.”

Paper Code MIT Tech Review CN

Adversarial Cooperative Rationalization: The Risk of Spurious Correlations in Even Clean Datasets

Wei Liu, Zhongyu Niu, Lang Gao, Zhiying Deng, Jun Wang, Haozhao Wang, and Ruixuan Li

“An interpretable, causal learning paradigm that simultaneously avoids spurious correlations in data and traditional self-interpretable models.”

Paper Code

👨‍🔧 Applications

MedTrinity-25M: A Large-scale Multimodal Dataset with Multigranular Annotations for Medicine

Yunfei Xie*, Ce Zhou*, Lang Gao*, Juncheng Wu*, Xianhang Li, Hong-Yu Zhou, Sheng Liu, Lei Xing, James Zou, Cihang Xie, and Yuyin Zhou (*: Joint first authors)

“A comprehensive, large-scale multimodal dataset for medical vision-language models.”

Paper Toolkit Code stars

Dataset Website Expert Eval

DyFlow: Dynamic Workflow Framework for Agentic Reasoning

Yanbo Wang, Zixiang Xu, Yue Huang, Xiangqi Wang, Zirui Song, Lang Gao, Chenxi Wang, Xiangru Tang, Yue Zhao, Arman Cohan, Xiangliang Zhang, and Xiuying Chen

“DyFlow is a dynamic workflow framework for LLM-based agents that adapts its reasoning steps in real time using intermediate feedback, enabling better generalization across diverse tasks.”

Paper Code

🧐 Service

2026 ACL ICML EMNLP NeurIPS TMLR HISS

2025 ACL NeurIPS EMNLP NLPCC ICLR HISS

💼 Experiences

[04 / 2026 - Now] ByteDance, Algorithm Intern
[02 / 2026 - 08 / 2026] Amazon, Remote Research Collaboration (topic: LLM Jailbreak & Information Security)
[10 / 2024 - 07 / 2025] MBZUAI, Research Intern (Supervisor: Dr. Xiuying Chen, topic: Mechanistic Interpretability of LLMs)
[07 / 2024 - 10 / 2024] University of Notre Dame, Research Intern (Supervisor: Prof. Xiangliang Zhang, topic: LLMs for Bayesian Optimization)
[01 / 2024 - 06 / 2024] UC Santa Cruz, Research Intern (Supervisor: Dr. Yuyin Zhou, topic: Vision-Language Models for healthcare)
[10 / 2023 - 12 / 2023] HUST (Supervisor: Prof. Ruixuan Li, topic: Interpretable deep learning frameworks)

💬 I am deeply grateful to all the mentors and collaborators who have guided and supported me along the way. Your encouragement, trust, and inspiration have made all the difference in my journey.

🏆 Honors and Awards

ACL Travel Grant, 2026
ICML Travel Grant, 2026
Outstanding Research Contribution in NLP Award (Student Award Category, ~top 2%), 2025
ACL Travel Grant, 2025

📖 Education

08 / 2025 - Now : Ph.D. student, Mohamed bin Zayed University of Artificial Intelligence

09 / 2021 - 07 / 2025 : B.E., Huazhong University of Science and Technology

🧩 Miscellaneous

📚 Resources

Insights

📖 Book Interpretability in Deep Learning 📖 Book Interpretable Machine Learning 📖 Book Trustworthy Machine Learning 📖 Book 大语言模型 (LLM Chinese) 📄 Article The Bitter Lesson 📄 Article The Urgency of Interpretability

Blogs

05/24 Mandarin National Undergraduate Innovation Project Documentation 03/24 Mandarin Negative Transfer 03/24 Mandarin Mixture of Experts Explained 01/24 Mandarin EMNLP2020 Tutorial Notes (Explainable AI)

Other Stuff

I also like photography. Sometimes I take good photos by accident. So I might upload a few here someday, along with some unnecessary commentary, but feel free to pretend you’re looking forward to it.🙃