Join our free seminar focusing on the second approach to the problem — speculative interpretability, as published by Anthropic and OpenAI on distill.pub. Despite somewhat shaky theoretical foundations and aggressive duck-typing, the results are striking. With some caveats — we must never forget the speculative assumptions involved — these techniques point toward a new generation of lightweight and accurate networks. The Qwen models appear to be the first moving in this direction.

About the Speaker
Raphael is a developer, educator and independent researcher in computer science and machine learning with over 20 years of consulting experience. He has worked across a wide range of industries and technical domains, from embedded optical sensors to international rating databases, spanning all layers of telecom systems. Throughout his career, he has used a broad variety of programming languages and frameworks. In recent years, his work has focused on scientific machine learning and formal descriptions of neural networks.

Date: Thursday, February 26, 2026
Location: Online
Cost: Free to attend

Register here: https://synteda.com/events/interpretability-for-llms/

SHARE