https://www.youtube.com/watch?v=0704iLc55Fs

Look at the papers

• Function Vectors in large language models
• Emergent World Representations
• Designing a Dashboard for transparency and control of conversational AI

Website to that does an interpretability on sparse autoencoders

https://www.neuronpedia.org/gemma-scope#main

Slides https://drive.google.com/file/d/1DqbVcNj-AMAecVJ_dpInV3Sx4-vL5cgG/view