AE-QSB algorithm achieves 74.6% best mean gap on benchmarks

Existing quantum-inspired simulated bifurcation (QSB) algorithms rely on static parameters and uniform strategies, leading to poor adaptability and homogenized results. To address this, Liu et al. introduce the Adaptive Enhanced Quantum-inspired Simulated Bifurcation (AE-QSB) framework, which leverages real-time perception of four distinct population states to create a closed-loop cycle of perception, decision-making, and execution. This allows the algorithm to dynamically adjust its search behavior based on the current solver state.

The framework includes three complementary algorithms: ME-BSB for efficient extremum seeking, SE-DSB for population-level uniform refinement, and SG-DSB for density-aware adaptive scheduling. On the medium-sized graph G22, SE-DSB and SG-DSB achieve a mean gap below 0.05%, while ME-BSB delivers the best trade-off between runtime and solution quality with a 0.26% gap and the shortest single-run time. Across all 81 benchmark graphs (G1–G81), AE-QSB variants achieve the lowest mean gap on 74.6% of graphs and the highest average approximation rate on 84.5%. Ablation experiments confirm that subgroup exploration and rescue mechanisms are critical to performance. This work demonstrates that statistical population information provides a computable foundation for adaptive control, transitioning quantum-inspired optimization from fixed scheduling to data-driven closed-loop control.