Lennart Finke

Hi! I work on language model evaluations and alignment, currently being a scholar at MATS mentored by the one and only Mary Phuong.

Before that, I was on exchange at Harvard, and before that, a summer fellow at Pivotal Research, examining language model societal impacts with the UK AI Security Institute and the OpenAI Preparedness Team. I've also contributed to METR's evaluations suite. Previously, I did a degree in math at the University of Göttingen.

🗞️ SimpleStories makes its way into Goodfire's Research Blog and Anthropic's Alignment Science Blog! Especially the latter one, on gradient routing, is a recommended read.

🇸🇬🇺🇸 Presented SimpleStories at the ICLR SynthData Workshop and NeurIPS 2025. Was nice seeing y'all in Singapore and San Diego!

📬 Anthropic Cofounder Jack Clark reports on our LLM persuasion paper. Thanks Jack!

Blog: fi-le.net
Email: firstname@finke.dev
Github: @lennart-finke
Twitter: @lennartfinke
HF: @lennart-finke
Papers: Scholar Page

If you want to have a call or get coffee in the Bay, especially if you're into AI safety, I'll probably say yes. Feel free to mail me!

The pixel art on the left is hand-drawn by me.

I really enjoy the Japanese language, holding JLPT N1 and 漢検２級.