Liver cancer remains one of the most severe diseases threatening human health globally. Early diagnosis plays an important role in improving the survival rate and reducing patient suffering. In clinical practice, multi-phase contrast-enhanced magnetic resonance imaging (MRI) can deliver more accurate liver lesion diagnosis, as there is more detailed visual information such as clearer blood vessels, mediastinum, and different image enhancement manners of various lesions. However, liver lesion diagnosis in 3D multi-phase images is a challenging task since liver lesion has variant shape, diverse size, irregular boundaries, and low signal-to-noise ratio. Meanwhile, each phase exhibits different image properties such as intensity and contrast, that is, each phase reflects different lesion characteristics. Moreover, to the best of our knowledge, there is no publicly available dataset of liver lesions with multi-phase MR scans, which limits the development of automatic computer-aided diagnosis (CAD) systems for multi-phase liver lesion identification. We plan to host the first Liver Lesion Diagnosis Challenge on Multi-phase MRI (termed LLD- MMRI2023) to promote the research of such CAD systems. In our challenge, 498 multi-phase MRI lesion cases are publicly released, which are carefully annotated by 3 experienced radiologists. Each case comprises 8 different phases, including the non-contrast phase, arterial phase, venous phase, delay phase, T2-weighted imaging phase, diffusion-weighted imaging phase, T1 in-phase, and T1 out-of-phase. We assigned 316cases for training, 78 cases for validation, and 104 cases for testing. The classification performance is reported with F1-score and Cohen's Kappa coefficient.