models:
  - model: rubricreward/R3-Qwen3-8B-LoRA-4k
    parameters:
      weight: 0.4
  - model: rubricreward/R3-Qwen3-8B-4k
    parameters:
      weight: 0.3
  - model: rubricreward/R3-Qwen3-8B-14k
    parameters:
      weight: 0.3
merge_method: ties
base_model: Qwen/Qwen3-8B
lambda: 1.0
dtype: bfloat16