ELLIS Institute Tübingen

AI Safety and Alignment Group

AI safety · alignment · evaluation

The AI Safety and Alignment Group is based at the ELLIS Institute Tübingen and the Max Planck Institute for Intelligent Systems. This Hub hosts our public datasets and trajectory releases for work on AI-agent evaluation, safety, and alignment.

Research projects

Project Paper Dataset
ResearchArena arXiv Trajectories
PostTrainBench arXiv Trajectories
InferenceBench arXiv Trajectories
Instrumental Choices arXiv Agent traces
Evaluation awareness arXiv EvalAwareBench

Browse all GitHub repositories, follow the group on Substack, or visit the group page.