arXiv:2610.29170
Turning Quantum Compiler Configuration Selection Upside Down: A Top-Down Paradigm
Ming Zhong, Xiangyu Ren, Kangyu Zheng, Zhi Zeng, Shenggen Zheng, Shaohua Li, Zhiding Liang
Abstract
Selecting an effective quantum compiler configuration requires coordinated decisions over both hardware targets and compiler settings, with the ultimate goal of improving circuit execution fidelity. Existing SOTA approaches typically decompose this problem into two sequential stages: a device is chosen first, after which compiler options are tuned on that device through RL-based search. This bottom-up strategy can be fundamentally limited by its initial device decision, since a poorly matched backend may constrain the best fidelity achievable regardless of how effectively later compiler options are optimized. In contrast to classical compiler tuning, where the search space is often prohibitively large, the configuration space of quantum compilers remains manageable enough to support exhaustive search. Consequently, rather than repeatedly searching from candidate configurations, one can derive reusable configuration knowledge from offline oracle results and apply it directly to new circuits. In this paper, we introduce a top-down configuration-selection paradigm and build QuOPTBench, an exhaustively searched benchmark containing oracle configurations for 268 quantum circuits. We first study whether LLMs can directly infer suitable configurations and evaluate ten representative LLMs. The results reveal that vanilla LLMs remain far from the oracle. We then employ an LLM to extract transferable rules from sampled circuits governing configuration selection. Using these rules together with predictors trained on the enumerated configuration data of the sampled circuits, we design QuTOP, a rule-guided ML framework that performs fast configuration selection without iterative online search. Across our evaluation, the distilled rules consistently strengthen LLM-based selection, while QuTOP provides the strongest overall balance between solution quality and selection latency among the baselines.
AI Summary
The paper studies joint selection of a quantum hardware backend and Qiskit compiler settings to improve execution fidelity without iterative online compilation. It proposes learning configuration-selection rules and predictors from offline exhaustive search, rather than selecting a device first and subsequently tuning compiler options. QuOPTBench enumerates up to 960 device–option combinations for each of 268 circuits across four IBM fake backends, requiring 8,345 CPU-hours. Its empirical oracles maximize a gate- and readout-error product, normalized by input operation count. Ten general-purpose LLMs generally underperform a baseline that chooses the capacity-feasible device with lowest average gate error and applies default $O_3$. From 68 sampled circuits, an LLM-assisted, manually checked workflow distills 13 selection rules. QuTOP combines these rules with four Extra Trees predictors for optimization opportunity, fidelity gain, compilation time, and timeout risk. On 200 held-out circuits, QuTOP reports mean headroom-normalized gains over the baseline of approximately zero, 3.47%, and 28.56% across three groups with increasing oracle improvement potential, selecting configurations in about 3.3 seconds. These percentages measure reduction of remaining estimated-fidelity error, not relative fidelity increases. Additional tests report 9.06% improvement on 100 new circuits with four new backends and 5.71% mean improvement using noisy-simulation Hellinger fidelity on small/medium circuits. Stated limitations include bounded oracle searches, reliance on estimated fidelity and fake-backend noise models, possible baseline-reimplementation differences, LLM nondeterminism, and researcher judgment in rule verification. Generalization across architectures, compiler versions, calibration drift, crosstalk, and real hardware remains unestablished.
Finally, we provide all case-level rules to Claude-Opus-5 and ask it to consolidate recurring patterns into a compact rule set, while also identifying the supporting cases for each distilled rule.
Error found: The abstract generalizes exhaustive-search feasibility beyond the restricted 960-configuration, fixed-seed, timeout-limited search actually evaluated. The noisy-simulation evaluation reports all 63 small and medium QASMBench circuits without reconciling this with the stated exclusion of rule-development circuits from evaluation.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 16 pages, 9 figures, 9 tables