-
12-10
2025
可计算性、结构与逻辑：Bakh Khoussainov教授特邀学术讲座
2025年12月9日，欧洲科学院院士、新西兰皇家学会会士、中国电子科技大学算法与逻辑实验室主任Bakh Khoussainov教授做客清华大学交叉信息院，带来题为“可计算性、结构与逻辑（Computability, structures, and logic）”特邀报告。交叉信息院院长姚期智教授主持讲座。师生齐聚FIT楼多功能厅，聆听讲座。精彩讲座姚先生首先介绍Bakh Khoussainov教授在数理逻辑与理论计算机科学领域的核心贡献。Khoussainov教授曾荣获欧洲理论计算...
-
12-26
2025
Rationality of multi-agent learning and uncertainty quantification for the generative models
演讲人：Yian Ma (马易安) [University of California, San Diego]时间：13:30-15:00, Dec 26, 2025 (Fri)地点：RM 1-222, FIT Building内容：In the first part of the talk, I will discuss an interesting phenomenon in multiagent learning, that the mixed Nash equilibria are uniformly stable if and only if they are collectively rational. This justifies the effusive use of multi-agent learning systems and r...
-
12-23
2025
祖冲之三超导量子计算进展
演讲人：龚明 [中国科学技术大学微尺度国家研究中心] 时间：15:00-17:30, Dec 23, 2025 (Tue)地点：RM S327, MMW Building内容：超导量子计算技术发展的核心目标是通过提升量子比特数目及操控质量，实现通用容错量子计算机。其实现路径上有三个重要的里程碑，分别是量子计算优越性、专用量子模拟、以及通用量子计算。本团队在2021年实现了量子优越性的里程碑，展示了超越经典计算的能力，为探索量子增强的近期应用提供了更多机...
-
12-23
2025
Pointwise Generalization in Deep Neural Networks
演讲人：Yunbei Xu [NUS]时间：11:00-12:00, Dec 23, 2025 (Tue)地点：RM 1-222, FIT Building内容：We address the fundamental question of why deep neural networks generalize by establishing a pointwise generalization theory for fully connected networks. This framework resolves long-standing barriers to characterizing the rich, nonlinear feature-learning regime and builds a new statistical foundation...
-
12-19
2025
Quantum Control and Measurement of Nuclear Spin Qubits in Diamond
演讲人：张钧翔 时间：13:30-16:00, Dec 19, 2025 (Fri)地点：RM S327, MMW Building内容：Nuclear spins ubiquitously present in diamond nitrogen-vacancy (NV) center systems exhibit exceptional properties as quantum resources, including long coherence times, efficient polarization, and versatile readout modalities. This report systematically investigates quantum control and measurement methodologies ...
-
12-16
2025
Digital Protection in the Era of Generative Models
演讲人：贺笛 [北京大学] 时间：12:00-13:00, Dec 16, 2025 (Tue)地点：FIT 1-312内容：随着大规模视觉生成模型的迅速扩展，内容来源可追溯性与模型滥用检测成为影响生成式 AI 健康发展的关键问题。模型级水印通过在生成过程中内嵌可识别信号，为判断图像是否由特定模型生成提供了必要的技术基础。然而，传统学习式水印依赖黑箱神经网络，其检测结果缺乏严格的可靠性保证，难以量化错误率。相比之下，具备统计可证性的水印方法...