quant-phAug 4, 2026

Separating quantum circuits from classical LLMs

Authors: Srinivasan ArunachalamArkopal DuttHari KroviRik Sengupta

Organizations: IBM Research

Abstract

Modern large language models - transformers and diffusion language models - are built around two canonical algorithmic tasks: prediction and generation. We prove unconditional separations between low-depth quantum computation and the corresponding bounded-resource classical language-model architectures in both regimes. Concretely, we exhibit the following: 1. Distributional separation. We give a distribution that is sampleable by QNC0\textsf{QNC}^0 circuits (i.e., a family of constant-depth quantum circuits consisting of bounded fan-in gates) that no constant-round diffusion language model (DLM\textsf{DLM}) with shallow scheduling and denoising can sample within constant distance, even when allowed sublinear chain-of-thought and output-token revision/remasking events, the very features modern DLM\textsf{DLM}s rely on. 2. Functional separation. We exhibit a function computable in QNC0[loglogn]\land \circ \textsf{QNC}^0[\log\log n] (i.e., a family of O(loglogn)(\log\log n)-depth QNC0\textsf{QNC}^0 circuits, where nn is the input length, followed by a single classical AND\mathsf{AND} gate) such that any constant-depth decoder-only transformer computing the function must be large: it would have to have width nΩ(1)n^{Ω(1)}. Together, our work initiates the study of quantum advantage in the era of large language models.

Explore similar work

CardsList