AI and the World's Hard Problems · Quantum physics & quantum computing · arXiv · 2026-10-19
Efficient LLM-Generated Shuttling Compilers for Complex Trapped-Ion Architectures
Shuttling compilers for trapped-ion quantum computers convert algorithms into sequences of ion movements. The authors had Claude Opus 4.7 generate and iteratively refine these compilers in Python from written specifications, and benchmarked them against hand-crafted compilers. The generated compilers cut shuttling timesteps by up to 72% for a linear trap and up to 43% for a trap with junctions, and a second model, Claude Fable 5, reproduced these findings.