The evaluation framework for knowledge-integrated AI utilizes a patient agent designed to simulate real patient behavior during multi-turn interactions with a candidate Doctor LLM, ensuring the agent is realistic enough to challenge the model while remaining deterministic for fair comparison.

Authors

• Person: Not available Organization: arXiv
  A Comprehensive Benchmark and Evaluation Framework for Multi ...

Sources

• A Comprehensive Benchmark and Evaluation Framework for Multi ... arxiv.org arXiv via serper

Referenced by nodes (1)

• Patient Agent concept