eval_results.json

{
    "eval_reasoning_loss": 0.6476022005081177,
    "eval_reasoning_runtime": 89.5729,
    "eval_reasoning_samples_per_second": 1.44,
    "eval_reasoning_steps_per_second": 0.368,
    "eval_samples": 100,
    "eval_utility_loss": 0.9602180123329163,
    "eval_utility_runtime": 33.9739,
    "eval_utility_samples_per_second": 1.442,
    "eval_utility_steps_per_second": 0.383
}