The Aloe Family Recipe for Open and Specialized Healthcare LLMs
Dario Garcia-Gasulla, Jordi Bayarri-Planas, Ashwin Kumar Gururajan, Enrique Lopez-Cuena, Adrian Tormos, Daniel Hinjos, Pablo Bernabeu-Perez, Anna Arias-Duart, Pablo Agustin Martin-Torres, Marta Gonzalez-Mallo, Sergio Alvarez-Napagao, Eduard Ayguadé-Parra, Ulises Cortés
2025-05-21
Summary
This paper talks about the Aloe Beta models, which are new open-source AI systems designed to give medical advice and information more accurately and safely.
What's the problem?
The problem is that most medical AI models either aren't open for everyone to use or they don't perform well enough to be trusted with important healthcare questions, and sometimes they can even give unsafe or unreliable answers.
What's the solution?
To solve this, the researchers improved how the AI is trained by cleaning up the data it learns from, teaching it to prefer better answers, letting it look up extra information when needed, and testing it very carefully to make sure it's both accurate and safe.
Why it matters?
This matters because it means doctors, patients, and researchers can use powerful, trustworthy medical AI tools that are open to everyone, helping improve healthcare for more people around the world.
Abstract
Aloe Beta models improve open-source medical Large Language Models through enhanced preprocessing, Direct Preference Optimization, Retrieval-Augmented Generation, and a rigorous evaluation methodology, achieving competitive performance and high safety standards.