reports/eval/base_a100_coding.json

{
  "label": "base_a100",
  "model": "Qwen/Qwen3.6-27B",
  "results": [
    {
      "kind": "humaneval",
      "n": 164,
      "pass@1": 0.939,
      "passed": 154
    }
  ],
  "reference_pass@1": {
    "Qwen2.5-Coder-32B-Instruct": 0.92,
    "GPT-4o": 0.9,
    "DeepSeek-V3": 0.91,
    "Llama-3.1-70B-Instruct": 0.8,
    "Qwen2.5-7B-Instruct": 0.84,
    "CodeLlama-34B": 0.51
  }
}