CorpusQA 1M

reasoning

CorpusQA 1M is a long-context question answering benchmark designed to evaluate models at approximately 1 million token contexts. Models are scored on accuracy when retrieving and reasoning over information distributed across an extremely long input corpus.