#multi-step reasoning'

'Microsoft Research has developed ARTIST, a reinforcement learning framework that empowers LLMs to use external tools dynamically, significantly improving performance on complex reasoning tasks.'