Tag
Looped language models enhance compositional tool calling by leveraging recurrent computation, improving accuracy on multi-step tasks while adaptive inference optimizes the balance between performance and compute cost. The study suggests these models are promising for reliable agentic systems.