eval_results.json

{
    "eval_reasoning_loss": 0.8993186950683594,
    "eval_reasoning_runtime": 29.1782,
    "eval_reasoning_samples_per_second": 4.421,
    "eval_reasoning_steps_per_second": 1.131,
    "eval_samples": 100,
    "eval_utility_loss": 1.4286295175552368,
    "eval_utility_runtime": 10.9929,
    "eval_utility_samples_per_second": 4.457,
    "eval_utility_steps_per_second": 1.183
}