-
08-13
2025
Generalizing from SIMPLE to HARD Visual Reasoning: Can We Mitigate Modality Imbalance in VLMs?
演讲人： 程韵 [普林斯顿大学] 时间： 11:00-12:00, Aug 13, 2025 (Wed)地点：RM 1-222, FIT Building (https://meeting.tencent.com/dm/hpLuprdKjM45 #腾讯会议：567-960-508)内容：While Vision Language Models (VLMs) are impressive in tasks such as visual question answering (VQA) and image captioning, their ability to apply multi-step reasoning to images has lagged, giving rise to perceptions of modality ...
-
08-11
2025
On the Power of Context-Enhanced Learning in LLMs
演讲人： 朱星宇 [普林斯顿大学]时间： 11:00-12:00, Aug 11, 2025 (Mon)地点：RM 1-222, FIT Building (https://meeting.tencent.com/dm/O15OBOE6VLfC #腾讯会议：490-402-711)内容：We formalize a new concept for LLMs, context-enhanced learning. It involves standard gradient-based learning on text except that the context is enhanced with additional data on which no auto-regressive gradients are computed. ...
-
07-30
2025
Goedel-Prover-V2: The Strongest Open-Source Theorem Prover to Date
演讲人： Chi Jin [Princeton University]时间： 11:00-12:00, Jul 30, 2025 (Wed)地点：RM 1-222, FIT Building (https://meeting.tencent.com/dm/wIj1OMGZx499)内容：This talk introduces Goedel-Prover-V2, an open-source model that establishes a new state-of-the-art for automated theorem proving in Lean. https://blog.goedel-prover.com/个人简介:Chi Jin is an Assistant Professor of Electrical and Computer ...
-
07-25
2025
New Breakthrough in Matrix Multiplication
演讲人： Renfei Zhou [CMU]时间： 11:00-12:00, Jul 25, 2025 (Fri)地点：RM 1-222, FIT Building (https://meeting.tencent.com/dm/1suGQN3X1y5V （#腾讯会议：939-634-954）)内容：Fast matrix multiplication is one of the most fundamental problems in computer science. We present new algorithms that improve the time complexity of matrix multiplication to $n^{2.371339}$, surpassing the previous bound of $n...
-
07-25
2025
量子电池容量理论及其应用
演讲人： 杨雪 [西南交通大学]时间： 10:00-11:00, Jul 25, 2025 (Fri)地点：RM S527, MMW Building内容：随着量子技术的发展，以可控方式存储和释放能量的量子设备（即量子电池）的理论和实验研究越来越重要。与传统的经典电池相比，量子电池具备更为快速且更高功率的充电性能，这一特性使其有望在降低人类对化石燃料的依赖、减少碳排放方面发挥重要作用。量子电池研究的一个核心问题是能量转换，此前的研究表明量子电池的能量...
-
07-24
2025
Selective conformal prediction for trusted AI-driven decisions
演讲人： 金滢 [宾夕法尼亚大学]时间： 10:30-12:00, Jul 24, 2025 (Thu)地点：RM 1-222, FIT Building内容：Conformal prediction is a distribution-free uncertainty quantification framework for assessing the reliability of black-box AI models. Standard conformal prediction provides on-average (marginal) guarantees which, despite being useful, can be insufficient in decision-making processes that usual...