-
把「An inverse theorem for the Gowers U^4 norm」写成可执行的阅读路径
Ben Green, Tamar Ziegler and I have just uploaded to the arXiv our paper…
-
2026元强化学习落地指南：核心原理、算法对比与任务生成全流程
元强化学习（Meta-RL）是强化学习与元学习的融合技术，核心价值是解决传统RL样本效率低、新场景适配慢、泛化能力弱的行业痛点。其通过双层优化架…
-
把「Course announcement: 254A – random matrices」写成可执行的阅读路径
Starting in the winter quarter (Monday Jan 4, to be precise), I will be …
-
把「Reading seminar 6: “Stable group theory and appr」写成可执行的阅读路径
This is the last reading seminar of this quarter for the Hrushovski pape…
-
「Non-commutative Freiman theorems, and model theo」：数学直觉如何落到可检验的步骤
This is an adaptation of a talk I gave recently for a program at IPAM. I…
-
2023 Transformer家族V2.0升级详解：解决长序列算力瓶颈的落地路径
Transformer V2.0是2023年对原生Transformer架构的全面重构升级，核心解决原生模型序列长度受限、算力二次方暴涨、静态建…
-
「Random Martingales and localization of maximal i」笔记：问题从哪来、证明卡在哪
Assaf Naor and I have just uploaded to the arXiv our paper “Random Marti…
-
把「Random covariance matrices: Universality of loca」写成可执行的阅读路径
Van Vu and I have just uploaded to the arXiv our paper “Random covarianc…
-
把「Approximate bases, sunflowers, and nonstandard a」写成可执行的阅读路径
One of the most basic theorems in linear algebra is that every finite-di…
-
2026大Transformer推理优化实战：从显存瓶颈到提速37倍全方案
大尺寸Transformer模型推理的核心瓶颈集中在显存占用过高、算力二次方增长、自回归推理并行度低三大问题，单纯依赖硬件扩容无法实现规模化落地…