AnalysisAI ModelsOctober 10, 2026

Podcast: Goodfire's Eric Bigelow on how LLMs decide

Read original source →youtube.com

Goodfire researcher Eric Bigelow joins The Cognitive Revolution to discuss mechanistic interpretability of model decision-making, covering his research into forking paths and critical tokens. He argues model reasoning works as in-context learning from sampled tokens.

People · Eric Bigelow

1 source

More stories today

Open the live feed