Unprivileged CUDA kernels can use Rowhammer to tamper with GPU page tables for targeted privilege escalation, leaking cryptographic keys and escalating to CPU root access by bypassing IOMMU.
super hub
ImageNet Large Scale Visual Recognition Challenge
52 Pith papers cite this work, alongside 30,004 external citations. Polarity classification is still indexing.
52
Pith papers citing it
30k
external citations · Crossref
hub tools
citation-role summary
background 2
dataset 2
citation-polarity summary
claims ledger
- dataset T able 4Common datasets used in CDOD benchmarks, summarizing modality, scale, annotation volume, typical role, and dominant shift type. Acronyms: S = Source, T = Target. Symbol:∼ indicates approximate counts. Dataset Y ear Modality #Images #Cls #Anno Role Domain Shift PASCAL VOC [95] 2007-2012 RGB∼16.5K∼20∼40K S/T mild scene shift MS COCO [96] 2014 RGB∼330K∼80∼2.5M S scene diversity ImageNet DET [97] 2013 RGB∼450K∼200∼500K S fine-grained cate- gory Cityscapes [98] 2016 RGB∼3.0K∼8∼65K T urban sce
- dataset Finally, ifg 1 andg 2 both do not depend on the second argument, (3) is a linear parabolic SPDE with additive noise: dUt =α 1(t)∆Ut dt+α 2(t) dWt for allt∈I.(20) I Numerical simulation For the numerical simulation of the forward and backward processes, (3) and (1), we modeled the image space Λ as Λ = (0, d1)×(0, d 2)and decomposed the boundary∂Λaccording to ∂LΛ :={0} ×[0, d 2);(21) ∂T Λ := [0, d1)× {d 2};(22) ∂RΛ :={d 1} ×(0, d 2];(23) ∂BΛ := (0, d1]× {0}(24) into its left, top, right and bottom
- background ImageNet Large Scale Visual Recognition Challenge.International Journal of Computer Vision (IJCV)115, 3 (2015), 211-252. doi:10.1007/s11263-015-0816-y [41] Shuran Song, Samuel P Lichtenberg, and Jianxiong Xiao. 2015. Sun rgb-d: A rgb-d scene understanding benchmark suite. InProceedings of the IEEE conference on computer vision and pattern recognition. 567-576. [42] Alex Tamkin, Mike Wu, and Noah D. Goodman. 2020. Viewmaker Networks: Learning Views for Unsupervised Representation Learning.ArXivab
- background 1 Introduction In recent years, the emergence and evolution of auto-regressive models [18, 44, 66] and diffusion models [32, 61, 16, 50, 58, 55, 56] have led to AI-generated content (AIGC) becoming increasingly realistic and widely applied across industries, bringing convenience to fields such as entertainment [51, 2, 63], advertising [ 39, 17], and medicine [ 60, 83]. This progress is particularly evident in AI- synthesized images, which have seen gradual improvements in resolution and semantic
authors
co-cited works
representative citing papers
State-of-the-art convolutional networks easily memorize random labels and unstructured noise images, indicating that generalization in deep learning cannot be explained by traditional capacity or regularization arguments.
HASTE enables training-free dynamic compression of pre-trained CNNs by patch-wise LSH-based merging of redundant channels, reporting 46.2% FLOPs reduction on ResNet34 CIFAR-10 with 1.25% accuracy drop.
Semantic geometry emerges transiently early in next-token prediction training before collapsing to Neural Collapse symmetry in synthetic settings with latent semantic factors.
DiTo shifts token reduction in DiTs to output token similarity, reusing prior-step matches across timesteps with PMR scheduling and frequency-aware penalties to raise PSNR at given speedups.
ImageAttributionBench is a benchmark dataset demonstrating that state-of-the-art image attribution methods lack robustness to image degradation and fail to generalize to semantically disjoint domains.
Single-thread JPEG benchmarks misrank decoders for ML DataLoader use, with rankings changing across CPUs and worker counts; torchvision and simplejpeg perform best in measured DataLoader tiers.
Variable codebook sizes that increase along the sequence in visual tokenizers reduce generation FID scores significantly for autoregressive models on ImageNet.
Multi-Level Optimal Transport (MOT) jointly infers soft layer couplings and neuron transport plans to produce global alignment scores and structured hierarchical correspondences between networks of varying depths.
ClusterMark applies visual token clustering to create robust in-generation watermarks for autoregressive image models, improving detectability under perturbations compared to direct token biasing while preserving quality.
SCOOTER supplies best-practice guidelines, open tools, and a 3K-image benchmark with 34K+ human ratings showing that six tested unrestricted attacks produce images humans can detect as fake.
LAION-5B is an openly released dataset of 5.85 billion CLIP-filtered image-text pairs that enables replication of foundational vision-language models.
SPN is a CNN that detects a spacecraft bounding box, classifies then regresses attitude, and optimizes position via Gauss-Newton, achieving degree-level attitude and cm-level position errors on real images after training only on synthetic data.
Mixed precision training uses FP16 for most computations, FP32 master weights for accumulation, and loss scaling to enable accurate training of large DNNs with halved memory usage.
FUSE creates full-spectrum unlearnable perturbations using random spectral masking during training and cross-band guidance to enforce consistency between frequency components.
ViPSy constructs policy-aligned and visually grounded preference pairs for VLMs via visual cues from image variants, yielding SOTA hallucination reductions of 35.7% on AMBER and 24.5% on Object HalBench.
RBFN projection heads serve as competitive replacements for MLP heads in SSL and enable SNS, a label-free metric from RBF parameters that correlates strongly with logistic regression evaluation.
Jaguar replaces prime-modulus HE with power-of-two arithmetic to enable coefficient-domain convolution and local-shift truncation, reporting 2-3.7x lower latency than Cheetah and Rhombus on ResNet-18/50 and MobileNetV2.
CSFlow derives inference-time timestep weights for flow matching by matching per-step frequency content to human CSF, yielding 4.7% FID reduction and smaller gains on IS and GenEval.
C-GSPN scales 2D spatial propagation to foundation vision encoders via a fast CUDA kernel, compressed blocks, and two-stage distillation, matching ViT performance with 15% fewer parameters and 4x block speedup at 2K resolution.
CS researchers show pragmatic skepticism toward LLM leaderboards, using them despite distrust while preferring peer networks, arena leaderboards, and cost transparency as key missing feature.
Introduces LOES, a constructive spectral method to select task-discriminative subspaces from intermediate layer embeddings, and GeoReg for enforcing simplicial class geometry during fine-tuning, with reported gains increasing with model depth across modalities.
Symmetrizing cross-entropy produces the unique convex multi-class unhinged loss, which locally approximates other symmetric losses, and enables new interpolating losses SGCE and alpha-MAE with competitive performance on noisy-label benchmarks.
HDFM adds a continuous heat-dissipation (blur) process to flow matching, aligns an interpolated path to fix ill-posed inverse heat dissipation, and uses x-prediction to ease high-dimensional regression, yielding better performance than most baselines on image datasets.
citing papers explorer
-
Mixed Precision Training
Mixed precision training uses FP16 for most computations, FP32 master weights for accumulation, and loss scaling to enable accurate training of large DNNs with halved memory usage.
-
EmergentBridge: Improving Zero-Shot Cross-Modal Transfer in Unified Multimodal Embedding Models
EmergentBridge enhances zero-shot cross-modal performance on unpaired modalities by learning noisy bridge anchors from existing alignments and enforcing proxy alignment only in the orthogonal subspace to avoid gradient interference.