{
  "model": "DeepSeek-Flash-Mini (nano)",
  "protocol": "loglikelihood multiple-choice, mean NLL argmin",
  "device": "cpu",
  "subjects": {
    "high_school_mathematics": {
      "correct": 84,
      "total": 270,
      "acc": 0.3111
    },
    "abstract_algebra": {
      "correct": 31,
      "total": 100,
      "acc": 0.31
    },
    "astronomy": {
      "correct": 28,
      "total": 152,
      "acc": 0.1842
    },
    "computer_security": {
      "correct": 17,
      "total": 100,
      "acc": 0.17
    },
    "management": {
      "correct": 29,
      "total": 103,
      "acc": 0.2816
    },
    "college_biology": {
      "correct": 29,
      "total": 144,
      "acc": 0.2014
    },
    "econometrics": {
      "correct": 28,
      "total": 114,
      "acc": 0.2456
    },
    "formal_logic": {
      "correct": 28,
      "total": 126,
      "acc": 0.2222
    },
    "global_facts": {
      "correct": 20,
      "total": 100,
      "acc": 0.2
    },
    "human_aging": {
      "correct": 48,
      "total": 223,
      "acc": 0.2152
    }
  },
  "overall_correct": 342,
  "overall_total": 1432,
  "overall_acc": 0.2388,
  "random_baseline": 0.25
}