config.json

{
  "base_model": "Dream-org/Dream-v0-Instruct-7B",
  "adapter_type": "LoRA plus reward head",
  "attention": "bidirectional",
  "pool_strategy": "mean",
  "mask_aware_pooling": true,
  "step_embedding": true,
  "lora": {
    "r": 16,
    "alpha": 32,
    "dropout": 0.05,
    "target_modules": [
      "q_proj",
      "v_proj"
    ]
  },
  "release_status": "public release",
  "checkpoint_identity_status": "exact submitted-paper main PRM",
  "paper_role": "process-reward model used by PRM Guided, Hybrid, and the snapshot diagnostics",
  "adapter_bytes": 36202468,
  "parameter_prefixes_kept": [
    "lora_A",
    "lora_B",
    "reward_head",
    "step_proj",
    "step_embed"
  ],
  "causal": false,
  "no_step_embed": false,
  "no_mask_aware": false,
  "lora_r": 16,
  "lora_alpha": 32,
  "lora_dropout": 0.05,
  "step_embed_dim": 256,
  "reward_hidden": 1024,
  "seed": 42,
  "max_steps": 3000,
  "best_step": 2500,
  "train_samples": 1276560,
  "validation_samples_sampled": 6000
}