LLMs as Kernel-Smoothers with Verifier Loops: interpolation plus execution enables limited reasoning | Raisolo