sft-qwen2.5-1.5b/eval-results/hendrycks_math/metrics.json

{
  "hendrycks_math": {
    "pass@1": {
      "num_entries": 5000,
      "avg_tokens": 638,
      "gen_seconds": 129,
      "symbolic_correct": 52.52,
      "no_answer": 5.26
    }
  }
}