Claas Beger
Santa Fe Institute
I am currently a Graduate Fellow at Santa Fe Institute, doing research on multimodal reasoning under Professor Melanie Mitchell. Prior to that I finished my Master’s in Computer Science at Cornell University, where I worked with Kevin Ellis, Kilian Weinberger and Saikat Dutta. My research interest centers broadly around Human-like Artificial Intelligence. For this purpose, I think it is the most promising direction to look towards Natural Intelligence, both with regard to the brain and psychology/cognition.
I am applying to PhD programs for the fall 2025 cycle.
news
| Mar 28, 2026 | I was awarded a fellowship by Princeton University’s Natural and Artificial Minds (NAM) Initiative |
|---|---|
| Feb 03, 2026 | New paper on using VLMs for visual concept hypothesis formation on Bongard problems out on arxiv! arXiv:2602.03038 |
| Feb 03, 2026 | New paper on a diverse evaluation benchmark for Code Generation Agents out on arXiv arXiv:2602.02262 |
| Oct 03, 2025 | New paper on abstraction capabilities across modalities on ARC tasks out on arXiv! Also check out Melanie’s blog: https://aiguide.substack.com/p/do-ai-reasoning-models-abstract-and |
| Sep 22, 2025 | “Investigating Abstraction Capabilities of the o3 Model Using Textual and Visual Modalities” accepted as a Spotlight at Neurips Workshop Multimodal Algorithmic Reasoning! |
selected publications
-
2025TLDR: An architecture of schema learning and iterative application can resolve arbitrarily deep compositional statements. -
2025TLDR: Decomposing multi-hop questions into single-step prolog definition improves performance on various long-context question datasets. -
2025TLDR: Various language models struggle with dry-execution of simple and advanced code structures (Recursion, Concurrency OOP) -
2025TLDR: Using improved clustering and a more diverse embedding approach our technique can more accurately compress preference datasets into human-readable constitutions -
-
2025TLDR: Vision-Language models have good performance on abstract reasoning tasks, but do not utilize the intended human-core knowledge priors.