Heuristic Rate-Adaptive Reconciliation in Continuous-Variable Quantum Key Distribution

In continuous-variable quantum key distribution (CV-QKD), the information reconciliation step plays a crucial role in ensuring the security and efficiency of the key exchange process. This paper investigates the use of heuristic and reinforcement learning-based methods to adaptively adjust the low-density parity check (LDPC) code rate in response to the time-varying channel signal-noise-ratio (SNR), with the goal of maximizing the final key rate. Leaving the SNR to fluctuate by ±1dB around a fixed value, we analyze the trade-off between the reconciliation efficiency and the frame error rate (FER). We propose a rate-adaptive reconciliation scheme that leverages reinforcement learning to select the optimal LDPC code rate, comparing its performance against traditional fixed-rate approaches. The results demonstrate that the proposed adaptive scheme significantly enhances the key rate and shows potential in applying for a wide range of SNR.