Neural Implicit Representations (INRs) have immense potential in image super-resolution tasks due to their ability to model continuously using coordinates. However, the high-dimensional complexity of hyperspectral image data and the lack of sufficient long-range modeling capabilities in INRs have hindered their development in hyperspectral image super-resolution (HSI-SR) tasks. To address these challenges, we propose a novel framework integrating Entropy-Driven Adaptive Clustering Attention (EDACA) within a continuous INR for HSI-SR. Our Entropy-Driven Clustering adaptively partitions features using information entropy, optimizing input for attention. The Clustering-Compressed Attention mechanism then efficiently processes these features, significantly reducing overhead while capturing long-range dependencies. A Cross-scale Feature Integration module leverages INR’s continuity to fuse spectral and spatial features with attention-derived context. Overall, our method achieves state-of-the-art performance on Harvard and Pavia Centre datasets, validated by extensive ablation studies.