Tilman Räuker
Tilman Räuker
Co-Director of pivotal-research.org. @safe_paper
Selected publications:
Paper: Transformers Use Causal World Models in Maze-Solving Tasks – accepted at ICLR 2025 Workshop on World Models
Paper: Structured World Representations in Maze-Solving Transformers – accepted at the UniReps Workshop, NeurIPS 2023
Paper: Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks – accepted at SaTML 2023
Previous things I did:
Led RFPs on Cybersecurity AI and Agent Evaluation @ AI Safety Fund
Technical AI safety Research Manager @ ERA 2023
Master Thesis at LUH AI: On Temporally-Extended Reinforcement Learning in Dynamic Algorithm Configuration (Theresa Eimer & Prof. Marius Lindauer)
As a hobby, I like designing prints.