Comment by sunir

Maybe. I use QWAN frequently when working with the coding agents. That requires an llm equivalent of interoception to recognize when the model understanding is scrambled or “aligned with itself” which is what qwan is.