About me
Hi! I’m Wei Jie Yeo, currently an independent AI safety researcher based in Singapore. My current interest are concentrated within the field of alignment research, spanning evaluations of eval-awareness and agentic misalignment, scalable alignment training, and alignment pretraining. Additionally, I am interested in investigating why alignment happens, e.g. conducting mechansitic experiments on inoculation prompting, or exploring the capability sink effect in modular pretraining. Given how current frontier models are approaching a point of potentially causing mass disruption, I believe this is a crucial time to be doing alignment research!
My research background started from natural language processing and mechanistic interpretability. My PhD research at Nanyang Technological University, supervised by Prof. Erik Cambria, focused on understanding language models through their explanations and internal mechanisms, including refusal, faithfulness, and robustness to adversarial attacks.
I am open to collaborations or opportunities in safety research! Feel free to reach out!
Employment
- Current: Independent AI Safety Researcher.
- March–September 2026: Machine Learning Engineer, TikTok.
- July–December 2025: AI Safety Research Intern, Huawei International.
- Spring 2025: MATS exploratory stage, working on mechanistic interpretability with Neel Nanda.
Selected Publications
Debiasing CLIP: Interpreting and Correcting Bias in Attention Heads
Wei Jie Yeo, Rui Mao, Moloud Abdar, Ranjan Satapathy, Erik Cambria
ICML, 2026.
[Paper] [Code]
Beyond I’m Sorry, I Can’t: Dissecting Large-Language-Model Refusal
Nirmalendu Prakash, Wei Jie Yeo, Amir Abdullah, Ranjan Satapathy, Erik Cambria, Roy Ka-Wei Lee
AAAI, Special Track on AI Alignment, 2026.
[Paper]
Understanding Refusal in Language Models with Sparse Autoencoders
Wei Jie Yeo, Nirmalendu Prakash, Clement Neo, Ranjan Satapathy, Roy Ka-Wei Lee, Erik Cambria
Findings of EMNLP, 2025.
[Paper] [Code]
Towards Faithful Natural Language Explanations: A Study Using Activation Patching in Large Language Models
Wei Jie Yeo, Ranjan Satapathy, Erik Cambria
EMNLP, 2025.
[Paper]
Mitigating Jailbreaks with Intent-Aware LLMs (Intent-FT)
Wei Jie Yeo, Ranjan Satapathy, Erik Cambria
Preprint, 2025.
[Paper] [Code]
A comprehensive review on financial explainable AI
Wei Jie Yeo, Wihan Van Der Heever, Rui Mao, Erik Cambria, Ranjan Satapathy, Gianmarco Mengaldo
Artificial Intelligence Review, 2025.
[Paper]
Self-training Large Language Models through Knowledge Detection
Wei Jie Yeo, Teddy Ferdinan, Przemyslaw Kazienko, Ranjan Satapathy, Erik Cambria
Findings of EMNLP, 2024.
[Paper] [Code]
Plausible Extractive Rationalization through Semi-Supervised Entailment Signal
Wei Jie Yeo, Ranjan Satapathy, Erik Cambria
Findings of ACL, 2024.
[Paper] [Code]
How Interpretable are Reasoning Explanations from Prompting Large Language Models?
Wei Jie Yeo, Ranjan Satapathy, Rick Siow Mong Goh, Erik Cambria
Findings of NAACL, 2024.
[Paper] [Code]