ImageNet
PulseAugur coverage of ImageNet — every cluster mentioning ImageNet across labs, papers, and developer communities, ranked by signal.
- used by CIFAR-10 70%
- used by Diffusion Models 70%
- instance of Diffusion Models 70%
- used by vision transformer 70%
- used by residual neural network 70%
- instance of Diffusion models of ion-channel gating and the origin of power-law distributions from single-channel recording 70%
- instance of magazine 70%
- used by ConvNeXt 70%
- instance of arXiv 60%
- instance of CIFAR-100 60%
- instance of CNNS 60%
- affiliated with arXiv 50%
13 天有情绪数据
-
Papers challenge deep learning theory with generalization bound critiques
Two papers, one from 2016 by Zhang et al. and another from 2019 by Nagarajan and Kolter, are discussed for their impact on deep learning theory. The 2016 paper demonstrated that standard neural networks could easily mem…
-
H-Sets framework uncovers feature interactions in image classifiers
Researchers have developed H-Sets, a new framework designed to uncover and attribute higher-order feature interactions within image classifiers. This method moves beyond analyzing individual features to understand how g…
-
New research explores sparse attention and multimodal reasoning for faster, more accurate AI
Researchers have developed novel methods to enhance reasoning capabilities in AI models, focusing on efficiency and accuracy. One approach, LessIsMore, introduces a training-free sparse attention mechanism that maintain…
-
新理论揭示监督学习中固有的几何盲点
研究人员发现监督学习中存在一个根本性的几何局限性，称为“几何盲点”。这一理论发现表明，标准的监督学习目标固有地保留了对标签相关方向的敏感性，即使这些方向与测试无关。这个盲点统一了几个已观察到的问题，包括非鲁棒特征、纹理偏差、损坏脆弱性和鲁棒性-准确性权衡。引入了一个新的诊断指标“轨迹偏差指数”（TDI）来衡量这种现象，并且提出的“PMH”方法在缓解这种现象方面显示出潜力。
-
OpenAI advances consistency models for faster, high-quality AI generation
OpenAI has introduced sCM, a new approach to continuous-time consistency models that significantly speeds up generative AI sampling. This method simplifies and stabilizes training, allowing models to generate high-quali…
-
机器学习视觉导览 (2015)
本资源集提供了机器学习的广泛概述，涵盖了从基础概念、视觉导览到理论基础和实际应用。它包括一个分类任务的视觉指南，对机器学习基准的科学和伦理的讨论，以及全面的教科书和课程材料的链接。此外，它还重点介绍了可解释机器学习的工具以及在生产环境中部署模型所需的工程实践。
-
CVPR panels to explore future of ML datasets and infrastructure
Two panels are scheduled to coincide with the CVPR conference, focusing on the future of datasets and next-generation ML infrastructure. The first panel, on data-centric approaches, will feature experts from ImageNet, H…
-
OpenAI and researchers reveal AI vulnerabilities to adversarial attacks
OpenAI researchers are exploring the transferability of adversarial robustness across different types of perturbations in neural networks. Their findings indicate that robustness against one perturbation type does not a…
-
大语言模型研究探索新的训练、评估和模型行为理解方法
研究人员正在开发新方法来提高大语言模型在各个领域的性能。一项研究介绍了 MemCoE，一个受认知启发的框架，用于大语言模型代理学习如何组织和更新长期用户记忆，从而增强个性化。另一篇论文 ReLay 探索了个性化大语言模型生成的摘要，发现虽然个性化提高了理解能力，但也引入了偏见和幻觉的风险。此外，一个名为 ClassEval-Pro 的新基准被创建，用于评估大语言模型在类级别代码生成方面的能力，揭示了当前前沿模型之间显著的性能差距。
-
Google AI 推出研究代理；OpenAI 详解网络训练和非线性计算
Google AI 推出了测试时扩散深度研究员 (TTD-DR)，这是一个模仿人类研究过程的新颖框架，通过迭代起草和修改报告来利用检索到的信息。该方法将报告撰写建模为一个扩散过程，通过搜索驱动的去噪机制来完善初稿。OpenAI 还发表了几篇论文，详细介绍了训练大型神经网络的技术，包括数据、流水线和张量并行，以及探索由于浮点运算导致的深度线性网络的非线性计算特性。此外，OpenAI 还讨论了深度学习的基础设施考虑因素以及一种称为权重归一…