Aikyam-Lab/CURE-MED-14B

CURE-MED-14B: Multilingual Medical Reasoning LLM

CURE-MED-14B is a 14 billion parameter large language model developed by Aikyam Lab and collaborators, specifically designed for multilingual medical reasoning. Built upon the Qwen/Qwen2.5-14B-Instruct model, it addresses the complexities of medical queries across diverse languages.

Key Capabilities & Features

• Multilingual Medical Reasoning: Specialized for open-ended medical questions in 13 languages, including Amharic, Yoruba, and Swahili.
• Curriculum-Informed Reinforcement Learning: Employs a unique approach integrating code-switching-aware supervised fine-tuning (SFT) and Group Relative Policy Optimization (GRPO) to improve logical correctness and language stability.
• Enhanced Performance: Designed to overcome challenges in multilingual medical reasoning, particularly in underrepresented languages.
• Robust Training & Evaluation: Trained and evaluated using CUREMED-BENCH, a high-quality multilingual benchmark with verifiable answers.

Good For

• Healthcare Applications: Ideal for medical reasoning tasks requiring high accuracy and language stability.
• Multilingual Support: Suitable for use cases needing to process medical information across a wide range of languages, including those with limited existing LLM support.
• Research & Development: Provides a strong foundation for further research into multilingual medical AI and reinforcement learning techniques.