CIFAR-10
A commonly used image classification benchmark dataset containing 32x32 color images across 10 classes, with 6,000 images per class and 60,000 in total. It is widely used for evaluating machine learning models.
Timeline (last 90 days)
The MNIST and CIFAR-10 datasets have become so common in tutorials that they lack discriminative value.
The paper proposes a drift calibration scheme: each tile uses a small number of probe reads to fit a 16-parameter field to recover programming accuracy; on CIFAR-10 from 1 hour to 1 year, digital correction is 2.1 to 13.3 percentage points better than the best cheap digital baseline, and in-array correction retains a gain of 0.6 to 3.7 points.
Through empirical studies on different architectures (such as fully connected networks, convolutional networks, and residual networks) and datasets (such as MNIST, CIFAR-10, and ImageNet subsets), the authors found that the training process continuously expands ray coverage, especially in directions where the Jacobian singular values are smaller.
Case studies on the thousand-seed CIFAR-10 model library and MultiBERTs validated the formula with real seed distributions, with error within 0.05 in that range.
The MYOPIA benchmark contains four decision trees generated by LLMs and refined manually, corresponding to the GTSRB, MNIST, CIFAR-10, and MSCOCO datasets respectively.
The paper tested GCD on CIFAR-10 and ImageNet, covering three architectures, ResNet-18, ResNet-50, and ViT-B/16, and evaluated four mainstream attack methods.
The paper evaluated RASS on four types of tasks: Gaussian mixture target distributions, FFHQ unconditional pixel-space generation, CIFAR-10 conditional generation, and latent-space diffusion based on Stable Diffusion 3.5.
RASS achieves approximately 20% speedup on CIFAR-10 under matched computational budget conditions.
The paper validates the method on two types of architectures—fully connected networks and Vision Transformers—using the MNIST and CIFAR-10 datasets.
This study is based on CIFAR-10, a relatively simple dataset, and the optimal rank may shift upward for more complex, higher-resolution tasks.
2 more timeline events
All Facts (12)
The MNIST and CIFAR-10 datasets have become so common in tutorials that they lack discriminative value.
50%UnverifiedThe paper proposes a drift calibration scheme: each tile uses a small number of probe reads to fit a 16-parameter field to recover programming accuracy; on CIFAR-10 from 1 hour to 1 year, digital correction is 2.1 to 13.3 percentage points better than the best cheap digital baseline, and in-array correction retains a gain of 0.6 to 3.7 points.
50%UnverifiedThrough empirical studies on different architectures (such as fully connected networks, convolutional networks, and residual networks) and datasets (such as MNIST, CIFAR-10, and ImageNet subsets), the authors found that the training process continuously expands ray coverage, especially in directions where the Jacobian singular values are smaller.
50%UnverifiedCase studies on the thousand-seed CIFAR-10 model library and MultiBERTs validated the formula with real seed distributions, with error within 0.05 in that range.
50%UnverifiedThe MYOPIA benchmark contains four decision trees generated by LLMs and refined manually, corresponding to the GTSRB, MNIST, CIFAR-10, and MSCOCO datasets respectively.
50%UnverifiedThe paper tested GCD on CIFAR-10 and ImageNet, covering three architectures, ResNet-18, ResNet-50, and ViT-B/16, and evaluated four mainstream attack methods.
50%UnverifiedThe paper evaluated RASS on four types of tasks: Gaussian mixture target distributions, FFHQ unconditional pixel-space generation, CIFAR-10 conditional generation, and latent-space diffusion based on Stable Diffusion 3.5.
50%UnverifiedRASS achieves approximately 20% speedup on CIFAR-10 under matched computational budget conditions.
50%UnverifiedThe paper validates the method on two types of architectures—fully connected networks and Vision Transformers—using the MNIST and CIFAR-10 datasets.
50%UnverifiedThis study is based on CIFAR-10, a relatively simple dataset, and the optimal rank may shift upward for more complex, higher-resolution tasks.
50%UnverifiedThe study uses a DDPM-architecture U-Net as the backbone on the CIFAR-10 dataset, testing five different rank values: {2, 4, 8, 16, 32}.
50%UnverifiedCIFAR-10 is an image classification benchmark dataset released by Alex Krizhevsky and Geoffrey Hinton in 2009, containing 60,000 32×32 pixel color images across 10 categories, with 50,000 for training and 10,000 for testing.
50%