Hyperspectral and multispectral image fusion is a widely used technique to generate high-resolution hyperspectral images (HR-HSI) by combining the spatial details of multispectral images (MSI) with the spectral richness of hyperspectral images (HSI). Existing methods, however, often rely on predefined degradation priors and fail to fully capture the complex spatial-spectral interactions, which limits their performance in practical scenarios. Therefore, this paper proposes an adaptive degradation-aware feature fusion framework to address these limitations.The proposed framework consists of three components: a feature fusion network and two degradation-aware networks for spatial and spectral domains (SpaDNet and SpeDNet). Firstly, a spatio-spectral cross-attention mechanism is introduced, which explores the hierarchical correlations between spatial and spectral features to enhance their interaction in the reconstructed images. Secondly, leveraging the physical characteristics of the degradation process, SpaDNet and SpeDNet adaptively learn degradation priors from the input images, enabling efficient blind fusion without predefined degradation information. Lastly, a subspace loss function is designed to decouple spatial domain interference during spectral degradation modeling, enhancing the precision of degradation-aware learning.The proposed framework is validated on benchmark datasets, including CAVE and Harvard, achieving improvements in peak signal-to-noise ratio (PSNR) by 0.81 dB and 0.76 dB compared to existing methods. In experiments conducted on a hybrid-resolution imaging system, the reconstructed spectral band images demonstrated superior structural texture and signal strength accuracy. These results underscore the proposed method's ability to produce HR-HSI with enhanced spatial-spectral consistency, even in real-world scenarios where degradation priors are unknown.The proposed adaptive degradation-aware framework introduces a novel approach to blind HSI-MSI fusion, addressing the limitations of existing methods. The integration of spatio-spectral cross-attention and adaptive degradation modeling ensures robust and accurate fusion results. Experimental results on both synthetic and real-world datasets demonstrate the method's effectiveness, offering a significant improvement over state-of-the-art methods.