Image de couverture de livre

DNNs May Determine Major Properties of Their Outputs Early, with Timing Possibly Driven by Bias

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Park, Song, Chun, Sanghyuk, Heo, Byeongho, Han, Dongyoon
Format:	Preprint
Publié:	2025
Sujets:	Machine Learning Computer Vision and Pattern Recognition
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Similarity of Neural Architectures using Adversarial Attack Transferability
par: Hwang, Jaehui, et autres
Publié: (2022)

Rotary Position Embedding for Vision Transformer
par: Heo, Byeongho, et autres
Publié: (2024)

Learning with Unmasked Tokens Drives Stronger Vision Learners
par: Kim, Taekyung, et autres
Publié: (2023)

DenseNets Reloaded: Paradigm Shift Beyond ResNets and ViTs
par: Kim, Donghyun, et autres
Publié: (2024)

Masking meets Supervision: A Strong Learning Alliance
par: Heo, Byeongho, et autres
Publié: (2023)

RL makes MLLMs see better than SFT
par: Song, Junha, et autres
Publié: (2025)

Match me if you can: Semi-Supervised Semantic Correspondence Learning with Unpaired Images
par: Kim, Jiwon, et autres
Publié: (2023)

Morphing Tokens Draw Strong Masked Image Models
par: Kim, Taekyung, et autres
Publié: (2023)

SeiT++: Masked Token Modeling Improves Storage-efficient Training
par: Lee, Minhyun, et autres
Publié: (2023)

Improved Probabilistic Image-Text Representations
par: Chun, Sanghyuk
Publié: (2023)

Multiplicity is an Inevitable and Inherent Challenge in Multimodal Learning
par: Chun, Sanghyuk
Publié: (2025)

Lipsum-FT: Robust Fine-Tuning of Zero-Shot Models Using Random Text Guidance
par: Nam, Giung, et autres
Publié: (2024)

Probabilistic Language-Image Pre-Training
par: Chun, Sanghyuk, et autres
Publié: (2024)

Token Bottleneck: One Token to Remember Dynamics
par: Kim, Taekyung, et autres
Publié: (2025)

When Test-Time Adaptation Meets Self-Supervised Models
par: Han, Jisu, et autres
Publié: (2025)

MaskRIS: Semantic Distortion-aware Data Augmentation for Referring Image Segmentation
par: Lee, Minhyun, et autres
Publié: (2024)

Scratching Visual Transformer's Back with Uniform Attention
par: Hyeon-Woo, Nam, et autres
Publié: (2022)

LongProLIP: A Probabilistic Vision-Language Model with Long Context Text
par: Chun, Sanghyuk, et autres
Publié: (2025)

Learning to See What You Need: Gaze Attention for Multimodal Large Language Models
par: Song, Junha, et autres
Publié: (2026)

Exploring Conditions for Diffusion models in Robotic Control
par: Shin, Heeseong, et autres
Publié: (2025)

On Background Bias of Post-Hoc Concept Embeddings in Computer Vision DNNs
par: Schwalbe, Gesina, et autres
Publié: (2025)

DaWin: Training-free Dynamic Weight Interpolation for Robust Adaptation
par: Oh, Changdae, et autres
Publié: (2024)

HYPE: Hyperbolic Entailment Filtering for Underspecified Images and Texts
par: Kim, Wonjae, et autres
Publié: (2024)

Model Stock: All we need is just a few fine-tuned models
par: Jang, Dong-Hwan, et autres
Publié: (2024)

BEEM: Boosting Performance of Early Exit DNNs using Multi-Exit Classifiers as Experts
par: Bajpai, Divya Jyoti, et autres
Publié: (2025)

Basis-Oriented Low-rank Transfer for Few-Shot and Test-Time Adaptation
par: Park, Junghwan, et autres
Publié: (2025)

Angular Gradient Sign Method: Uncovering Vulnerabilities in Hyperbolic Networks
par: Jo, Minsoo, et autres
Publié: (2025)

I-SplitEE: Image classification in Split Computing DNNs with Early Exits
par: Bajpai, Divya Jyoti, et autres
Publié: (2024)

Evaluating Reliability in Medical DNNs: A Critical Analysis of Feature and Confidence-Based OOD Detection
par: Anthony, Harry, et autres
Publié: (2024)

TinyTrain: Resource-Aware Task-Adaptive Sparse Training of DNNs at the Data-Scarce Edge
par: Kwon, Young D., et autres
Publié: (2023)

Identifying Spurious Biases Early in Training through the Lens of Simplicity Bias
par: Yang, Yu, et autres
Publié: (2023)

Defining and Extracting generalizable interaction primitives from DNNs
par: Chen, Lu, et autres
Publié: (2024)

Magic for the Age of Quantized DNNs
par: Sawada, Yoshihide, et autres
Publié: (2024)

Exploring the Camera Bias of Person Re-identification
par: Song, Myungseo, et autres
Publié: (2025)

Mitigating the Bias in the Model for Continual Test-Time Adaptation
par: Chung, Inseop, et autres
Publié: (2024)

Motion-Oriented Compositional Neural Radiance Fields for Monocular Dynamic Human Modeling
par: Kim, Jaehyeok, et autres
Publié: (2024)

Technical Note: Defining and Quantifying AND-OR Interactions for Faithful and Concise Explanation of DNNs
par: Li, Mingjie, et autres
Publié: (2023)

Attack Anything: Blind DNNs via Universal Background Adversarial Attack
par: Lian, Jiawei, et autres
Publié: (2024)

Mix-and-Match Pruning: Globally Guided Layer-Wise Sparsification of DNNs
par: Monachan, Danial, et autres
Publié: (2026)

Efficient Noise Mitigation for Enhancing Inference Accuracy in DNNs on Mixed-Signal Accelerators
par: Azizi, Seyedarmin, et autres
Publié: (2024)