refactor: Task3 reward model changed, agent adjusted for new model

refactor: Improved grading logic for task 2

refactor: grader trivial bug,

Structure Changed, files reviewed

task3 evalutated

task1, task2 evaluated