Tag
论文提出 LSA-PAST，一种面向组合式增量学习（composition-IL）的参数高效方法，通过在冻结 ViT 中注入深度特化的 LoRA 适配器并用无复习的仿射变换将旧原型映射到新适配器空间，实现了精度-遗忘权衡的 state-of-the-art 表现。
该论文提出 EGiME（熵引导模型评估），一种基于特征空间聚类结构的半监督准则，无需分类器训练即可在少样本场景下评估预训练模型，在五个基准上与真实准确率相关性达 0.966，并比有监督基线提供更稳定的模型排名。
QuadTok introduces a hierarchical quadtree-based visual tokenizer that dynamically allocates tokens to intricate image regions, saving ~10% of tokens on ImageNet while enabling a 947M GPT-style model to reach 2.08 gFID on ImageNet 256x256 and zero-shot spatially controlled generation.
LingBot-Vision is a family of self-supervised vision transformer backbones for dense spatial perception, using masked boundary modeling to capture semantic and geometric structures for tasks like depth estimation and segmentation.
UniSpace introduces a unified visual representation that unifies semantic understanding, high-fidelity reconstruction, and image generation in a single space using a reparameterized ViT, eliminating the need for a separate VAE.
This paper presents a comprehensive comparison of deep learning architectures, including Vision Transformers and Graph Attention Networks, for automated sleep apnea detection from multichannel EEG signals, achieving a best test AUC of 0.750 using a vision transformer trained on topological data analysis features.
Introduces qZACH-ViT, a quantization-aware extension of ZACH-ViT with recursive intrinsic explanations, and Recursive Attribution-Stabilized Optimization (RASO) for stable attribution gradients. Achieves high prediction agreement and speedups on MedMNIST datasets after INT8 conversion.
This paper proposes a hardware-software co-design framework for N:M sparse vision Transformer inference, achieving over 2.2× latency speedup on GPUs while maintaining accuracy through a novel CUDA kernel (MD-SpMM) and a deployment-aware sparsity search.
Introduces Sparse-Reslim, a plug-in routing module that processes only 25% of spatial tokens through expensive transformer blocks for efficient weather prediction, achieving up to 3.18x speedup and improved forecast accuracy.
This paper proposes a token-centric dual-view learning framework that unifies prompt-based adaptation and cross-view fusion within a frozen vision transformer to improve breast cancer classification from mammography images, achieving consistent improvements on VinDr-Mammo and CMMD datasets.
LingBot-Vision introduces masked boundary modeling for self-supervised pretraining, achieving a 0.296 RMSE on NYUv2 linear-probe with 1.1B parameters versus 0.309 for DINOv3-7B, though it trails on ImageNet; weights are released in four sizes.
Patch-PODiff-ViT introduces a structured latent diffusion framework using patchwise Proper Orthogonal Decomposition (POD) for super-resolution and uncertainty quantification, enabling efficient diffusion with a fixed linear orthonormal basis and analytic propagation of predictive variance.
MuSViT is the first foundation vision model for sheet music, pre-trained on millions of pages via Masked Autoencoders, achieving superior performance in score recognition and symbol detection tasks.
Trained a superhuman Generals.io agent using self-play reinforcement learning with a JAX-based pipeline and Vision Transformer. Achieved #1 on human 1v1 leaderboard; all code and a fast JAX simulator open-sourced.
HYDRA-X presents a unified multimodal model that integrates image and video tokenization within a single Vision Transformer, achieving strong performance across understanding and generation tasks.
This paper proposes a method to convert pretrained Softmax attention models into linear-complexity Test-Time Training (TTT) architectures, achieving comparable text-to-image quality to fine-tuned Softmax models while significantly accelerating inference. The approach is validated by linearizing Stable Diffusion 3.5, resulting in SD3.5-T^5 with 1.32x speedup at 1K resolution.
A Rust crate for building and experimenting with Vision Transformer (ViT) models, providing typed configs, reusable structs, and runnable examples for research and production.
This paper proposes a new architecture that augments Flux Neural Operators with recurrent Vision Transformers to solve conservation laws as a foundation model. It demonstrates robust generalization and long-time prediction capabilities across diverse conservative systems without explicit access to governing equations.
This is a model card for OpenAI's CLIP model, detailing its Vision Transformer architecture, zero-shot image classification capabilities, and intended use for research purposes.
Deployment of Meta's DINOv3 ViT-L/16 distilled model on Replicate, providing efficient visual embeddings for tasks like similarity search and retrieval.