Qwen3-4B-ReMeDi

This repository contains the publicly released model weights of Qwen3-4B-ReMeDi, a Qwen3-4B model fine-tuned for multi-turn medical dialogue generation.

The model is released for the paper:

Experience is the Teacher: Reusing Atomic Thoughts from LLMs to Improve Medical Dialogue
ACL Findings 2026

Intended Use

This model is intended for research use in medical dialogue generation, multi-turn consultation modeling, medical reasoning, and evaluation of LLM-based healthcare dialogue systems.

Training Data

The model is trained on multi-turn medical dialogue data derived from the ReMeDi setting and further enhanced using the proposed Atomic Thoughts reuse framework.