CoRPS-Dial
A medical dialogue synthesis framework based on a dual-agent architecture that decouples patient personas from the clinical workflow, assigning them to a patient agent and a doctor agent respectively, with post-hoc quality control. It is used to generate clinically compliant and diverse pre-consultation dialogue data.
Timeline (last 90 days)
CoRPS-Dial is a dual-agent medical dialogue synthesis framework from an arXiv paper that decouples patient personas from the clinical workflow, carried respectively by a patient agent and a doctor agent.
In CoRPS-Dial, the patient agent is driven by structured personas, and the persona parameters are independent of the clinical workflow, allowing pairing with different workflows to generate diverse dialogue samples.
In CoRPS-Dial, the doctor agent advances the dialogue following guideline-aligned clinical workflow rules, ensuring that key information points are collected one by one.
CoRPS-Dial performs post-hoc quality control on the generated dialogues, checking three dimensions: role consistency, workflow completeness and text consistency, and unqualified samples are filtered out.
The paper validated CoRPS-Dial only in the coronary heart disease scenario, and the cost of cross-department transfer remains unclear.
After fine-tuning models on the data generated by CoRPS-Dial, performance on medical dialogue tasks validated the practical utility of the synthetic data.
All Facts (6)
CoRPS-Dial is a dual-agent medical dialogue synthesis framework from an arXiv paper that decouples patient personas from the clinical workflow, carried respectively by a patient agent and a doctor agent.
50%UnverifiedIn CoRPS-Dial, the patient agent is driven by structured personas, and the persona parameters are independent of the clinical workflow, allowing pairing with different workflows to generate diverse dialogue samples.
50%UnverifiedIn CoRPS-Dial, the doctor agent advances the dialogue following guideline-aligned clinical workflow rules, ensuring that key information points are collected one by one.
50%UnverifiedCoRPS-Dial performs post-hoc quality control on the generated dialogues, checking three dimensions: role consistency, workflow completeness and text consistency, and unqualified samples are filtered out.
50%UnverifiedThe paper validated CoRPS-Dial only in the coronary heart disease scenario, and the cost of cross-department transfer remains unclear.
50%UnverifiedAfter fine-tuning models on the data generated by CoRPS-Dial, performance on medical dialogue tasks validated the practical utility of the synthetic data.
50%