Multi-worm tracking plays a critical role in behavioral analysis, giving it broad applications in biomedicine and neuroscience. However, challenges such as small object sizes, high visual similarity, and frequent occlusions complicate reliable tracking. To address these, we propose MWT-HC, a new Multi-Worm Tracking framework that leverages Hierarchical Cues to enhance worm detection and tracking accuracy in complex microenvironments. The key insights of MWT-HC are as follows. First, we develop a dual-task detector for joint object detection and pose estimation. Within the detector, a hybrid-sensitive label assignment strategy and an occlusion-aware attention mechanism are introduced to improve the perception of small worms under occlusions. Second, we introduce a hierarchical cue-based tracker that integrates global cues for contextual understanding and local cues for structural detail preservation. Experiments on WormTrackV2 show that MWT-HC significantly improves tracking performance, with HOTA increased by 17.9%, AssA improved by 4.1 points, corresponding to a 12.6% gain, and IDF1 boosted by 9.6 points, yielding a 19.7% improvement over the leading multi-object tracking method. Ablation studies further validate the effectiveness of each module.