What is an LLM really thinking?
Very interesting articles about the way LLMs think in the background (without explicitly telling you in the chain of thought). The J-lens is an interpretability breakthrough that reveals what a model "thinks but doesn't say." Please read following interesting articles for details: