Google Deepmind proposes ‘self-discover’ framework for LLMs, improves GPT-4 performance

In 21 out of 25 tasks, the self-discover LLM framework was found to be outperforming chain-of-thought reasoning and other techniques with performance gains of up to 32%.

ai business computer science computers & electronics conversational ai deepmind found framework generative-ai google google brain google deepmind gpt gpt-4 gpt-4 performance large language models llm llm reasoning llm research llms ml and deep learning nlp performance reasoning science tasks thought university of southern california