由相似图构建加权独立集，平衡样本质量与多样性，为高效数据选择提供新框架。
arXiv:2605.15691v1 Announce Type: new Abstract: Data selection seeks to identify a compact yet informative subset from large-scale training corpora, b…
不规则时间序列生成预训练新方法ITGPT，突破传统模型在多模态缺失数据上的局限。
arXiv:2605.16069v1 Announce Type: new Abstract: Timeseries regression models often struggle to leverage large volumes of labeled multimodal data, part…
提出熵自编码器框架，通过隐式自由能最小化解决VAE后验崩溃，理论创新显著。
arXiv:2605.16164v1 Announce Type: new Abstract: Despite their ubiquity, variational autoencoders (VAEs) inherently suffer from posterior collapse, a f…
基于广义Onsager原理，提出假设驱动的介观动力学建模新范式，摆脱传统方程依赖，开启复杂多尺度系统建模新思路。
arXiv:2605.16211v1 Announce Type: new Abstract: Traditional scientific modeling typically begins with fixed, instance-wise effective equations and the…
AI视频推理新突破：用时空提示增强第一人称视频理解，还附带中间步骤评估，填补了基准空白。
arXiv:2605.15342v1 Announce Type: cross Abstract: Video reasoning models are a core component of egocentric and embodied agents. However, standard ben…
用语义级奖励替代二元反馈，让LLM学会表达真实不确定性，提升高安全场景下的可靠性。
arXiv:2605.15588v1 Announce Type: cross Abstract: As large language models (LLMs) are deployed in consequential settings such as medical question answ…
研究揭示大模型能否掌握未见编程语言：仅凭语法而非语义，LLM在全新语言上表现如何？
arXiv:2605.15607v1 Announce Type: cross Abstract: Large language models (LLMs) achieve high pass rates on code generation benchmarks, yet whether they…
离线数据下的风险感知策略学习新框架，用悲观原则优化高风险场景的决策效果
arXiv:2605.15620v1 Announce Type: cross Abstract: We study risk-aware offline policy learning, aiming to learn a decision rule from logged data that i…
医学指代图像分割遇上半监督学习，跨模态对齐降低标注成本，精准融合文本与视觉。
arXiv:2605.15720v1 Announce Type: cross Abstract: Medical referring image segmentation (MRIS) requires pixel-level masks aligned with textual descript…
用可控扩散模型生成容量约束点画，告别传统慢速迭代优化，实现高效图像条件点阵生成
arXiv:2605.15816v1 Announce Type: cross Abstract: Stipple patterns, point sets whose local density tracks a target image, are traditionally produced b…
首次给出非对数凹分布采样相对Fisher信息保证的查询复杂度，基于近端采样器与限制高斯oracle，是采样理论的重要进展
arXiv:2605.15859v1 Announce Type: cross Abstract: We study the query complexity of obtaining a relative Fisher information guarantee for sampling from…
机器学习新工具：无监督检测域迁移，还能定位异常特征子空间，实现可解释归因。
arXiv:2605.15920v1 Announce Type: cross Abstract: We developed a tool for detecting domain shifts, namely subtle differences in the probability distri…
探讨机器学习决策系统中可解释AI的不足，提出算法可争议性新视角，挑战传统XAI框架。
arXiv:2605.16041v1 Announce Type: cross Abstract: Machine learning systems increasingly make life-changing decisions about individuals, such as loan a…
无时钟异步电路驱动的可扩展神经形态计算架构，在FPGA上实现布尔脉冲神经元网络，突破传统时钟同步限制。
arXiv:2605.16114v1 Announce Type: cross Abstract: We propose a scalable neuromorphic architecture based on spiking dynamics emerging from the autonomo…
二阶优化方法加速LLM训练的瓶颈被Asteria运行时系统破解，大幅提升训练效率。
arXiv:2605.16184v1 Announce Type: cross Abstract: Second-order methods offer an attractive path toward more sample-efficient LLM training, but their p…
探索KST替代公式作为神经网络设计新范式，挑战经典数学定理的深度学习实践
arXiv:2410.01990v3 Announce Type: replace Abstract: This paper explores alternative formulations of the Kolmogorov Superposition Theorem (KST) as a fo…
低资源硬件感知NAS：仅需10次延迟探测就能高效搜索网络架构，降低对精确延迟模型的依赖
arXiv:2504.00663v2 Announce Type: replace Abstract: Existing hardware-aware NAS (HW-NAS) methods typically assume access to precise information circa …
简单kNN方法竟击败复杂学习路由器，重新思考LLM路由预测建模的惊人发现！
arXiv:2505.12601v2 Announce Type: replace Abstract: As large language models (LLMs) grow in scale and specialization, routing--selecting the best mode…
全新方法利用MoE正交生长，大幅节省LLM预训练成本，突破沉没成本陷阱。
arXiv:2510.08008v2 Announce Type: replace Abstract: As the computational demands for pre-training Large Language Models (LLMs) continue to surge, the …
量子机器学习的不确定性量化难题，本文提出自适应共形预测框架解决可靠性问题。
arXiv:2511.18225v2 Announce Type: replace Abstract: Quantum machine learning seeks to leverage quantum computers to improve upon classical machine lea…