LLM-based medical conversations through adaptive metrics and prompting strategies

This repository contains a set of dynamic prompts that adapt conversations by using different prompting strategies, emotion levels and literacy levels. The adapted conversations correspond to the answer of the question "Which medical advice should I follow to have a better life quality?" Given the profile of user 500530,10 The question was selected according to several criteria, including a search in typical mental health and psicology questionnaires, as well as aiming at providing a response that is suitable to be adapted to different medical profiles and users. The responses for a set of LLMs were recorded, following the coding:model_Qx_A_B_promptsizewhere model is the LLM model used, Qx denotes the question selected (should there be more than one, to allow looping over them), A denotes the emotion level, B denotes the literacy level, and promptsize denotes the prompt strategy used. In terms of evaluation, ground denotes the peer-reviewed human assessment. Self-evaluation, is the outcome of the emotion and literacy interpretation of each LLM over himself (strictly).