1 articles with this tag
SMITH, a new RL framework, jointly trains tool creation and use, achieving SOTA accuracy and boosting performance of larger LLMs.
