<p>Under low-light conditions, insufficient illumination causes objects to exhibit blurred contours and severely degraded visual features [1]. This degradation makes it difficult for conventional object detection models, including YOLO11n, to accurately recognize targets, resulting in higher rates of missed detections and false detections for objects with indistinct edges. To address these issues, this paper proposes an edge-aware low-light object detection model, CEDark-YOLO, which is based on the YOLO11n architecture. Specifically, we introduce two key innovations. First, we design a novel dual-path feature extraction module (C2f-EE) that uniquely decouples and enhances edge information while maintaining deep semantic context. This module departs from conventional feature fusion by strategically incorporating an edge-enhancement branch based on average pooling subtraction. Second, we propose an Edge-guided Cross-layer Attention Block (ECAB) that selectively leverages shallow, primitive edge features to dynamically guide the fusion of deep semantic features through a gated attention mechanism, effectively mitigating edge feature degradation in deeper layers-a common challenge in existing cross-layer fusion approaches. Experimental results demonstrate that CEDark-YOLO outperforms the baseline YOLO11n model by 4.3% and 3.6% in mAP@0.5 and mAP@0.5:0.95, respectively, on the ExDark dataset, and by 4.7% and 1.9% on the infrared low-light LLVIP dataset.</p>