with a tool, an LLM agent can easily perform this task. If it evaluates its performance against a bunch of different tools and then learns, this is learnable. j

with a tool, an LLM agent can easily perform this task. If it evaluates its performance against a bunch of different tools and then learns, this is learnable. just not with pure LLM approach