Ilia Shumailov

Auditing LLM Token Usage and Extracting Reasoning Traces

Researchers have developed methods to extract reasoning traces from proprietary large language models, allowing for a deeper understanding of their decision-making processes. Separately, a developer created a token audi…