微博反腐是中国制度反腐的重要补充。近年来，随着被微博曝光的腐败问 题复杂性增加，信息失真更容易发生，以致民众的负面情绪更容易爆发。这对政府的腐 败治理提出了新的挑战与要求，但是学界对微博反腐的新特征仍然认识不足。以纵向约 束理论为视角，并利用大数据分析方法来深度剖析“北极鲶鱼”案例，可以揭示微博反腐 的新特征。在微博反腐的初期，地方政府的及时回应能够有效降低议题热度；但是进入 中期，地方政府的模糊回应会激发民众的不满情绪与非理性评论，而中央媒体则积极引 导民众互动；进入尾期，中央媒体与民众互动加强，可以促进地方政府回应。在一体推进 不敢腐、不能腐、不想腐的大背景下，各级政府应根据微博反腐的新特征，积极处理微博 反腐的负面舆情，推动国家与网络社会之间的良性互动。

Cyber anti-corruption, an important supplementary anti-corruption channel in China, has recently developed new features. The problem exposed is increasingly complex and multifaceted. Netizens become more polarized, and it is easy to trigger social resentment. While cyber anti-corruption has been a new challenge for the government, little is known about its new development and features. We employ a vertical accountability perspective and a machine-learning approach to analyze big data from Weibo texts, by analyzing the Beiji Nianyu-case. Local government′s quick responses can relieve citizens′ concerns and attention during the opinion expression period. Nevertheless, in the second stage, local government′s implicit responses may trigger citizens′ dissatisfaction and irrational behaviors, while central media may proactively interact with citizens. Central media strengthen the interactions with citizens at the peak, increasing local government′s responsiveness. It uncovers the complicated interactions of the central and local governments and citizens, and provides practical implications for local governments to enhance their governance in cyber anti-corruption.