Method evaluate

AgentBench.old/src/tasks/composite_task.py:32–39

(self, agent: Agent)

            self.tasks.append(sub_task)

    def evaluate(self, agent: Agent) -> Dict[str, Dict[str, Any]]:
        print(f"Evaluating Composite Task '{self.name}' ...")
        results = {}
        for task in self.tasks:
            result = task.evaluate(agent)
            results[task.name] = result
        self.save_metrics_all(results)
        return results

    def _sub_output_dir(self, sub_task: Task) -> str:
        target_category = sub_task.category or sub_task.name or "default"