{
  "method": "logistic regression on pairwise feature differences",
  "training_set_size": 1717,
  "feature_names": [
    "word_count",
    "char_count",
    "ttr",
    "sentence_count",
    "avg_sentence_len",
    "dialogue_ratio",
    "action_ratio",
    "exclamation_per_1k",
    "ellipsis_per_1k",
    "comma_per_1k",
    "avg_word_len",
    "bigram_repetition"
  ],
  "cv_accuracy": {
    "mean": 0.5387,
    "std": 0.0201
  },
  "feature_coefficients": {
    "word_count": 0.038616634397416805,
    "char_count": 0.03864279473772371,
    "ttr": -0.01914308086445501,
    "sentence_count": 0.028155957104233048,
    "avg_sentence_len": -0.0036095904495603687,
    "dialogue_ratio": 0.0437777895165803,
    "action_ratio": -0.006029425140718956,
    "exclamation_per_1k": 0.006058409728573959,
    "ellipsis_per_1k": -0.01170504339618912,
    "comma_per_1k": -0.0025946344729306414,
    "avg_word_len": 0.004955500028980793,
    "bigram_repetition": 0.017184690623146537
  },
  "intercept": -0.06435372402856071,
  "validation_against_human_arena": {
    "n_models_overlap": 11,
    "rho_vs_human_arena_elo": 0.5545,
    "p_value": 0.0767,
    "interpretation": "moderate proxy"
  },
  "per_model": {
    "mistral_small_creative": {
      "engagement_skill_mean": 0.0064,
      "engagement_skill_std": 0.0943,
      "n_scored": 27
    },
    "deepseek_r1_0528": {
      "engagement_skill_mean": -0.0108,
      "engagement_skill_std": 0.1449,
      "n_scored": 27
    },
    "claude_opus_4_6": {
      "engagement_skill_mean": -0.0265,
      "engagement_skill_std": 0.1829,
      "n_scored": 27
    },
    "gemini_3_1_pro": {
      "engagement_skill_mean": -0.0289,
      "engagement_skill_std": 0.1024,
      "n_scored": 25
    },
    "claude_opus_4_7": {
      "engagement_skill_mean": -0.0296,
      "engagement_skill_std": 0.1154,
      "n_scored": 27
    },
    "gemma_4_26b": {
      "engagement_skill_mean": -0.031,
      "engagement_skill_std": 0.1359,
      "n_scored": 27
    },
    "qwen3_5_flash": {
      "engagement_skill_mean": -0.0343,
      "engagement_skill_std": 0.0904,
      "n_scored": 25
    },
    "glm_4_7": {
      "engagement_skill_mean": -0.0544,
      "engagement_skill_std": 0.0937,
      "n_scored": 27
    },
    "claude_sonnet_4_5": {
      "engagement_skill_mean": -0.0623,
      "engagement_skill_std": 0.1219,
      "n_scored": 27
    },
    "gemini_3_1_flash_lite": {
      "engagement_skill_mean": -0.0647,
      "engagement_skill_std": 0.0652,
      "n_scored": 27
    },
    "kimi_k2_5": {
      "engagement_skill_mean": -0.0757,
      "engagement_skill_std": 0.0769,
      "n_scored": 27
    },
    "deepseek_v3_2": {
      "engagement_skill_mean": -0.0804,
      "engagement_skill_std": 0.1485,
      "n_scored": 27
    },
    "gemini_2_5_flash": {
      "engagement_skill_mean": -0.0812,
      "engagement_skill_std": 0.0873,
      "n_scored": 27
    },
    "kimi_k2_6": {
      "engagement_skill_mean": -0.0872,
      "engagement_skill_std": 0.0898,
      "n_scored": 23
    },
    "deepseek_v4_pro": {
      "engagement_skill_mean": -0.0949,
      "engagement_skill_std": 0.106,
      "n_scored": 27
    },
    "minimax_m2_7": {
      "engagement_skill_mean": -0.0994,
      "engagement_skill_std": 0.1583,
      "n_scored": 27
    },
    "grok_4_1": {
      "engagement_skill_mean": -0.1055,
      "engagement_skill_std": 0.1303,
      "n_scored": 27
    },
    "llama_4_maverick": {
      "engagement_skill_mean": -0.1059,
      "engagement_skill_std": 0.0723,
      "n_scored": 27
    },
    "glm_5_1": {
      "engagement_skill_mean": -0.1159,
      "engagement_skill_std": 0.1023,
      "n_scored": 26
    },
    "deepseek_v4_flash": {
      "engagement_skill_mean": -0.1186,
      "engagement_skill_std": 0.0867,
      "n_scored": 27
    },
    "gpt_4_1": {
      "engagement_skill_mean": -0.1388,
      "engagement_skill_std": 0.0625,
      "n_scored": 27
    }
  }
}