PhD Candidate · AI Safety

Tibo Vanleke

Researching LLMs, interpretability, and technical AI safety at Vrije Universiteit Brussel.

I’m currently exploring subliminal learning and how to make model behavior more legible, trustworthy, and aligned with real-world constraints.

LLMs + Safety Primary research focus
VUB Vrije Universiteit Brussel

Now

Building a research portfolio, experimenting with interpretability tools, and sketching a weekly writing habit for sharing ideas in public.

Research Threads

Subliminal Learning

What subtle signals teach models, and how to detect them early.

LLM Safety

Robustness, alignment, and safeguards for real deployments.

Interpretability

Tools and frameworks to illuminate model internals.

PhD Supervision

Prof. Vincent Ginis

Main promotor · Data Analytics Labs, Vrije Universiteit Brussel.

Prof. Filip Van Droogenbroeck

Co-promotor · Data Analytics Lab in Business Technology & Operations, Faculty of Social Sciences & Solvay Business School (VUB).

Prof. Marie-Anne Guerry

Co-promotor · Faculty of Social Sciences & Solvay Business School (VUB).

Selected Work

Curriculum Vitae

Publications, talks, and academic background.

VUB Data Lab

Research department and lab homepage.

Google Scholar

Citation list and research outputs.

Projects

Experiments, tools, and open-source work.

Blog

Research Notes
A weekly memo format for safety and interpretability insights.
Bootstrapping a weekly research memo
February 4, 2026 · Proof-of-concept post.
Reading notes: interpretability papers
Planned • A weekly digest of papers and takeaways.