gaia_benchmark/run.py

from questions import get_all_questions
from gaia_interface import submit_to_leaderboard  # Adjust path if needed

def run_and_submit_all(agent):
    questions = get_all_questions()
    answers = []
    for q in questions:
        answer = agent(q)  # Pass entire dict, not q["question"]
        answers.append({"question_id": q["id"], "answer": answer})
    return submit_to_leaderboard(answers)