Low-Light Image Enhancement (LLIE) aims to restore details and visual information in images affected by low-light conditions, thereby improving their visibility and quality. While numerous deep learning approaches based on convolutional neural networks (CNNs) and Transformers have achieved remarkable progress in LLIE, existing hybrid methods that combine CNNs for local feature extraction and Transformers for global dependency modeling face inherent issue. Specifically, Transformers tend to suffer from interference caused by irrelevant regions when computing self-attention, and this limitation persists in hybrid approaches, impacting reconstruction quality. To address this issue, we propose a Hybrid Adaptive Sparse Cross-Attention Network (HASNet), which utilizes a novel lightweight Adaptive Sparse Cross-Attention Block (ASCAB) based on Transformer with a Channel Attention Block (CAB). Additionally, a Dual-scale Gated Feed-Forward Network (DGFFN) is introduced to further enhance low-light image features. Extensive experiments demonstrate that our method achieves superior performance across multiple benchmarks, validating the effectiveness and practicality of HASNet in LLIE.