5 ms·
Beyond the Black Box: Interpretability of LLMs in Finance
- ashater 1y agoOur 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.