Yes but at that point you may as well use a proper hyperparameter tuning framework like optuna if all the LLM agent is supposed to do is do hyperparameter tuning.
Does optuna think abstractly (i.e. use LLM to interpret the code and come up with insights), or just perform hyperparameter tuning experiments on user-indicated parameters?
Does optuna think abstractly (i.e. use LLM to interpret the code and come up with insights), or just perform hyperparameter tuning experiments on user-indicated parameters?