CDR Dataset
A dataset for evaluating contextual default reasoning and belief retraction in language models.
Projects
A collection of selected research projects and open-source work, including datasets, benchmarks, and tools developed to explore and evaluate AI reasoning and behavior.
SymRet Dataset
A dataset for evaluating rule propagation, belief retraction, and counterfactual rule propagation on symbolic graphs.
Multi-Agent Dark-Web Safety
A synthetic testbed for studying how privacy violations emerge and spread through multi-agent collaboration.
