math_with_judge:
  resources_servers:
    math_with_judge:
      entrypoint: app.py
      judge_model_server:
        type: responses_api_models
        name: policy_model
      judge_responses_create_params: {
        input: []
      }
      should_use_judge: false
      domain: math
      verified: false
      description: Hermes Agent with terminal, file, code_execution, skills, todo toolsets on OpenMathReasoning math dataset with math-verify and LLM-as-a-judge
      value: Improve model math capabilities in hermes agent harness such as AIME25

math_with_judge_hermes_agent:
  responses_api_agents:
    hermes_agent:
      entrypoint: app.py
      resources_server:
        type: resources_servers
        name: math_with_judge
      model_server:
        type: responses_api_models
        name: policy_model
      enabled_toolsets: [terminal, file, code_execution, skills, todo]
      max_turns: 30
      concurrency: 32
      terminal_backend: local
      terminal_timeout: 30
      system_prompt: |
        You are a precise math assistant. You have access to a terminal tool to run Python for calculations or verification.
        Use it when computation would help. Always state your final answer as \boxed{answer} in your response.
      datasets:
      - name: example
        type: example
        jsonl_fpath: resources_servers/math_with_judge/data/example.jsonl
      - name: train
        type: train
        jsonl_fpath: resources_servers/math_with_judge/data/OpenMathReasoning_train.jsonl
        huggingface_identifier:
          repo_id: nvidia/Nemotron-RL-math-OpenMathReasoning
          artifact_fpath: train.jsonl
        license: Creative Commons Attribution 4.0 International