Comment by ashater

Our paper introduces AI explainability methods, mechanistic interpretation, and novel Finance-specific use cases. Using Sparse Autoencoders, we zoom into LLM internals and highlight Finance-related features. We provide examples of using interpretability methods to enhance sentiment scoring, detect model bias, and improve trading applications.