Description
Sharpness-Aware Minimization, or SAM, is a procedure that improves model generalization by simultaneously minimizing loss value and loss sharpness. SAM functions by seeking parameters that lie in neighborhoods having uniformly low loss value (rather than parameters that only themselves have low loss value).
Papers Using This Method
SAMO: A Lightweight Sharpness-Aware Approach for Multi-Task Optimization with Joint Global-Local Perturbation2025-07-10Sharpness-Aware Machine Unlearning2025-06-16From Sharpness to Better Generalization for Speech Deepfake Detection2025-06-13Towards Understanding The Calibration Benefits of Sharpness-Aware Minimization2025-05-29An Analysis of Concept Bottleneck Models: Measuring, Understanding, and Mitigating the Impact of Noisy Annotations2025-05-22Improving Generalization of Medical Image Registration Foundation Model2025-05-10Learning from Loss Landscape: Generalizable Mixed-Precision Quantization via Adaptive Sharpness-Aware Gradient Aligning2025-05-08Focal-SAM: Focal Sharpness-Aware Minimization for Long-Tailed Classification2025-05-03Federated EndoViT: Pretraining Vision Transformers via Federated Learning on Endoscopic Image Collections2025-04-23Mitigating Parameter Interference in Model Merging via Sharpness-Aware Fine-Tuning2025-04-20DGSAM: Domain Generalization via Individual Sharpness-Aware Minimization2025-03-30FairSAM: Fair Classification on Corrupted Data Through Sharpness-Aware Minimization2025-03-29The Devil is in Low-Level Features for Cross-Domain Few-Shot Segmentation2025-03-27Understanding Flatness in Generative Models: Its Role and Benefits2025-03-14Training Diagonal Linear Networks with Stochastic Sharpness-Aware Minimization2025-03-14Effective LLM Knowledge Learning via Model Generalization2025-03-05LORENZA: Enhancing Generalization in Low-Rank Gradient LLM Training via Efficient Zeroth-Order Adaptive SAM2025-02-26Monge SAM: Robust Reparameterization-Invariant Sharpness-Aware Minimization Based on Loss Geometry2025-02-12Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyond2025-02-07GCSAM: Gradient Centralized Sharpness Aware Minimization2025-01-20