simpleqa:
  resources_servers:
    simpleqa:
      entrypoint: app.py
      domain: knowledge
      verified: false
      description: SimpleQA short-form factual QA with 3-tier LLM judge (CORRECT / INCORRECT / NOT_ATTEMPTED)
      value: Evaluate factual recall and abstention calibration via LLM-graded open-ended QA

      judge_model_server:
        type: responses_api_models
        name: genrm_model
      judge_responses_create_params:
        input: []
        max_output_tokens: 2048
        temperature: 0.0
        top_p: 1.0
      use_chat_completions_for_judge: true

simpleqa_simple_agent:
  responses_api_agents:
    simple_agent:
      entrypoint: app.py
      resources_server:
        type: resources_servers
        name: simpleqa
      model_server:
        type: responses_api_models
        name: policy_model
      datasets:
      - name: example
        type: example
        jsonl_fpath: resources_servers/simpleqa/data/example.jsonl
        license: MIT