利用GPT-5等前沿大模型驱动自动机器学习，实现跨语言手写OCR的闭环神经架构搜索，突破传统方法瓶颈。
arXiv:2607.15509v1 Announce Type: cross Abstract: We present a fully automated closed-loop AutoML framework that uses GPT-5, GPT-4o, and Claude Sonnet…
从同步理论窥见注意力的本质：完美同步无法计算，FSN通过“受阻同步”实现结构化差异，为Transformer架构提供全新数学视角。
arXiv:2606.18694v1 Announce Type: new Abstract: A network of oscillators that synchronizes perfectly computes nothing further, so an attention archite…
ICML 2026 新研究，探索让 LLM 在推理时跳过或循环特定层，实现更高效、更智能的层分配策略。
arXiv:2606.06574v1 Announce Type: new Abstract: Large language models (LLMs) perform inference by following a fixed depth and order, non-recurrent exe…
引入LITT架构，通过虚拟相对时间轴对齐临床事件序列，实现事件时序注意力与个性化轨迹解读
arXiv:2602.10385v4 Announce Type: replace-cross Abstract: Automatically discovering personalized trajectories (i.e., sequential event patterns) from l…
针对KAN网络参数爆炸与高频特征捕捉难题，提出基于傅里叶变换的改进架构，理论与实验俱佳
arXiv:2502.06018v3 Announce Type: replace-cross Abstract: Although Kolmogorov-Arnold-based interpretable networks (KANs) possess strong theoretical ex…
突破性研究：首次实现无限深和宽Transformer的可训练性，彻底解决深层网络训练瓶颈
arXiv:2605.17660v1 Announce Type: cross Abstract: Transformers have become the dominant architecture in modern machine learning, yet the theoretical u…
探讨组合稀疏性作为神经网络架构设计的归纳偏置，从理论上破解高维学习中的维度灾难。
arXiv:2605.14764v1 Announce Type: cross Abstract: Identifying the structural priors that enable Deep Neural Networks (DNNs) to overcome the curse of d…