当足够接近还不够时：量子电路合成中的自回归漂移
When Close Enough Is Not Enough: Autoregressive Drift in Quantum Circuit Synthesis
浏览论文内容
中文总结 AI 辅助
研究容错计算的量子电路优化问题，用带结构化电路令牌化的变压器评估。在参数化电路上混合方法有效，在 Clifford + T 电路上存在自回归漂移致精确等效性下降，推理时策略和数据缩放可缓解但仍有问题。
中文摘要 AI 辅助
容错计算的量子电路优化需要精确的功能等效性，同时尽量减少昂贵的非 Clifford 资源（如 T 门）。我们使用具有结构化电路令牌化的紧凑 4480 万参数编码器 - 解码器变压器来研究此问题，在参数化电路（2 - 6 量子比特）和 Clifford + T 电路（3 - 6 量子比特）上进行评估。在参数化电路上，一种混合方法（变压器的结构，经典优化的角度）在 3 - 6 量子比特电路上实现了中位数保真度 1.000。在 Clifford + T 电路上，模型学习有效的语法和准确的 T 计数统计，但随着目标长度增加，精确等效性急剧下降。我们将此失败归因于自回归漂移，早期令牌差异通过从左到右的解码不可恢复地级联。两种方法部分缓解了漂移，推理时策略将精确匹配率从 7%提高到 22.5%，训练数据增加 2.5 倍将其提高到 39.5%。但目标长度增加时的下降仍然存在。设置之间的对比是我们的核心发现：当可以通过后处理挽救近似输出时，变压器成功；当需要精确离散正确性时，自回归漂移限制可靠性，推理时搜索和数据缩放是有效手段，而训练端微调与模型级多样化则不然。
英文摘要
Quantum circuit optimization for fault-tolerant computing requires exact functional equivalence while minimizing expensive non-Clifford resources such as T gates. We study this problem using a compact 44.8M-parameter encoder-decoder transformer with structured circuit tokenization, evaluating on parameterized circuits (2-6 qubits) and Clifford+T circuits (3-6 qubits). On parameterized circuits, a hybrid approach -- structure from the transformer, angles from classical optimization -- achieves median fidelity 1.000 on 3-6 qubit circuits. On Clifford+T circuits, where all gates are discrete and no post-processing is possible, the model learns valid syntax and accurate T-Count statistics, yet exact equivalence degrades sharply with target length -- from 88% on circuits with <=9 gates to near zero beyond 26 gates. We trace this failure to autoregressive drift: early-token divergence cascading irrecoverably through left-to-right decoding. Two levers partially mitigate the drift: inference-time strategies that generate multiple candidates and select via equivalence verification raise exact-match rates from 7% to 22.5%, while scaling training data by 2.5x pushes them to 39.5%. Yet the degradation with target length persists -- even with more data, exact equivalence drops from 94% on short circuits to under 4% beyond 26 gates. The contrast between settings is our central finding: when approximate outputs can be rescued by post-processing, the transformer succeeds; when exact discrete correctness is required, autoregressive drift limits reliability, with both inference-time search and data scaling as effective levers while training-side fine-tuning and model-level diversification are not.
