capability_head_config.json

{
  "format_version": 1,
  "head_type": "capability",
  "base_model": "google/gemma-4-E4B-it",
  "model_family": "gemma4",
  "thinking_mode": "off",
  "variant": "full-trajectory",
  "architecture": {
    "num_labels": 1,
    "head_input_mode": "completion_text_only",
    "hidden_encoder_type": "lite",
    "selected_hidden_layer_indices": null,
    "hidden_layer_selection": "last",
    "hidden_layer_index": null,
    "hidden_layer_indices": null,
    "max_seq_len": 32000,
    "max_pixels": 200000,
    "dtype": "bf16",
    "aux_label_column": "correctness_score",
    "regression_loss": "mse",
    "failure_threshold": 0.5
  },
  "weight_file": "capability_head.pt",
  "weight_sha256": "624f4cf4a8df7ca686fbc710dad6262f43c910ed46a7a1a931b7a7357d1fc3aa",
  "parameter_count": 2863517,
  "paper_use": "Table 2 routing and Table 4 web-search control",
  "code": "https://github.com/Amirhosein-gh98/Multi-Head-Latent-Control",
  "paper": "https://arxiv.org/abs/2607.14277"
}