On: Self-reasoning tokens: teaching models to think ahead

[refulgentis]

I'll call it "not even wrong" :P here, they're putting it in the model, you're describing a common bit of working with LLMs across memory / RAG / etc.