Recently, Tencent AI Lab, in collaboration with several academic institutions both domestically and internationally, has published a comprehensive review paper on the issue of large model hallucination. The research indicates that, compared to traditional models, large model hallucination assessment faces new challenges such as large data scale, high versatility, and being less noticeable. To reduce hallucinations, interventions can be implemented at various stages including pre-training, fine-tuning, and reinforcement learning. However, further research is needed on reliable evaluation methods to promote the practical application of large models.