mrcr_resources_server:
  resources_servers:
    mrcr:
      entrypoint: app.py
      domain: other
      verified: false
      description: Multi-round coreference resolution over multi-turn conversations with prefix-gated SequenceMatcher grading
      value: Improve long-context in-context retrieval and needle-count-aware reasoning

mrcr_simple_agent:
  responses_api_agents:
    simple_agent:
      entrypoint: app.py
      resources_server:
        type: resources_servers
        name: mrcr_resources_server
      model_server:
        type: responses_api_models
        name: policy_model
      datasets:
      - name: example
        type: example
        jsonl_fpath: resources_servers/mrcr/data/example.jsonl
        num_repeats: 1
        license: MIT