{
  "dataset": "humaneval",
  "n": 4,
  "model": "base",
  "model_id": "Qwen/Qwen3.5-4B",
  "model_revision": "851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a",
  "pass_at_1": 1.0,
  "passed": 4,
  "total": 4,
  "per_problem": [
    {
      "task_id": "HumanEval/0",
      "passed": true,
      "parse_ok": true,
      "n_answer_tokens": 188,
      "truncated": false
    },
    {
      "task_id": "HumanEval/1",
      "passed": true,
      "parse_ok": true,
      "n_answer_tokens": 214,
      "truncated": false
    },
    {
      "task_id": "HumanEval/2",
      "passed": true,
      "parse_ok": true,
      "n_answer_tokens": 98,
      "truncated": false
    },
    {
      "task_id": "HumanEval/3",
      "passed": true,
      "parse_ok": true,
      "n_answer_tokens": 158,
      "truncated": false
    }
  ],
  "greedy": true,
  "seed": 0,
  "thinking": "off",
  "max_new_tokens": 512,
  "visible_tests": 1,
  "timeout_s": 5.0,
  "mbpp_config": null,
  "offset": 0,
  "n_parse_failed": 0,
  "n_truncated": 0,
  "runner_sha256": "cf1a766a70df8e3aacae18fc52a1a2186f2f8fc20161a6ef679b3dd51f60fcd2",
  "resolved_sampling": {
    "temperature": 0.0,
    "top_p": 1.0,
    "top_k": 0,
    "min_p": 0.0,
    "presence_penalty": 0.0,
    "frequency_penalty": 0.0,
    "repetition_penalty": 1.0
  },
  "generation_seconds": 5.9841226630378515,
  "sampled_tokens_per_second": 110.62607457714351,
  "model_load_seconds": 111.95083298604004,
  "wall_seconds": 133.71673452993855
}