A team of researchers from the HLTCOE and CLSP were named winners of an Outstanding Paper Award at the inaugural Conference on Language Modeling held in Philadelphia in October, 2024. The paper was called “Dated Data: Tracing Knowledge Cutoffs in Large Language Models” and investigates the concept of “reported knowledge cutoffs” commonly provided by large language models (LLMs) and we find that the knowledge in some LLMs do not match their reported cutoffs. The full paper can be found here.