EVALUATING LARGE LANGUAGE MODELS ON LONG TEXTS

Summary from the TLDR AI newsletter

Ada-LEval is a new benchmark designed to rigorously test large language models on their ability to understand long and ultra-long documents.