Function main

scripts/eval_arena_hard_auto_rtc.py:172–185

()

    logger.info(f"RTC pass rate: {passed_rtc_count / total_examples * 100:.2f}%")

def main():
    parser = argparse.ArgumentParser(description="Evaluate LLMs on arena-hard-auto dataset using RTC")
    parser.add_argument("--model", type=str, required=True, help="OpenAI model to use")
    parser.add_argument("--output", type=str, default="rtc_eval_results.json", 
                      help="Output file for results")
    
    args = parser.parse_args()
    
    # Create results directory if it doesn't exist
    os.makedirs("results", exist_ok=True)
    output_file = os.path.join("results", args.output)
    
    # Run evaluation
    evaluate_dataset(args.model, output_file)

if __name__ == "__main__":
    main()