LLM-Driven Algorithm Design for Quantum Circuit Synthesis based on Binary Decision Diagrams
QuantumEvo uses an LLM to generate BDD ordering heuristics, then keeps the ones that lower quantum circuit cost.
The paper targets reversible circuit synthesis, where BDD variable ordering can strongly affect the final quantum cost. Its method does not ask the LLM to output an ordering directly; it searches over heuristic programs and selects them by downstream QCC. The discovered heuristic, HGA-QE, changes the sifting step inside a genetic algorithm to better track that cost. On the benchmark set, it ties or beats the per-function best baseline 70.9% of the time and is strictly best on 13.5% of functions. ArXiv · AI/CL/LG's note
The paper targets reversible circuit synthesis, where BDD variable ordering can strongly affect the final quantum cost. Its method does not ask the LLM to output an ordering directly; it searches over heuristic programs and selects them by downstream QCC. The discovered heuristic, HGA-QE, changes the sifting step inside a genetic algorithm to better track that cost. On the benchmark set, it ties or beats the per-function best baseline 70.9% of the time and is strictly best on 13.5% of functions. ArXiv · AI/CL/LG's note
score 4