Researchers find you don’t need a ton of data to train LLMs for reasoning tasks https://t.co/8vveMIO8EL https://t.co/ROX7G3594O on https://twitter.com/AcerboLivio/status/1890521404749000784