An X-Ray Is Worth 15 Features: Sparse Autoencoders for Interpretable Radiology Report Generation

no code implementations • 4 Oct 2024 • Ahmed Abdulaal, Hugo Fry, Nina Montaña-Brown, Ayodeji Ijishakin, Jack Gao, Stephanie Hyland, Daniel C. Alexander, Daniel C. Castro

Using an off-the-shelf language model, we distil ground-truth reports into radiological descriptions for each SAE feature, which we then compile into a full report for each image, eliminating the need for fine-tuning large models for this task.

Language Modelling Multimodal Reasoning