Results 81 to 90 of about 5,833,461 (242)

UniUltra: Interactive Parameter-Efficient SAM2 for Universal Ultrasound Segmentation

open access: yes
The Segment Anything Model 2 (SAM2) demonstrates remarkable universal segmentation capabilities on natural images. However, its performance on ultrasound images is significantly degraded due to domain disparities.
Chen, Xin   +9 more
core  

Cosmological fine-tuning arguments what (if anything) should we infer from the fine-tuning of our universe for life?

open access: yes, 2020
"If the physical constants, initial conditions, or laws of nature in our universe had been even slightly different, then the evolution of life would have been impossible.
Waller, Jason 1980-
core  

Structure‐Guided Dynamical Mechanics in Bioinspired Double Twisted Laminated Thin Films

open access: yesAdvanced Functional Materials, EarlyView.
This work explores a dynamic mechanistic study of crack propagation and puncture testing via real‐time monitoring of crack initiation, propagation, and termination through nature‐inspired single and double twisted architectures fabricated from cellulose nanocrystals.
Justin Brackenridge   +4 more
wiley   +1 more source

Optimizer-Aware Fine-Tuning of Whisper Small with Low-Rank Adaption: An Empirical Study of Adam and AdamW

open access: yesInformation
Whisper is a transformer-based multilingual model that has illustrated state-of-the-art behavior in numerous languages. However, the efficiency remains persistent with the limited computational resources.
Hadia Arshad   +5 more
doaj   +1 more source

Electrospun Wood‐Derived Biopolymers as Electrodes in Electrochemical Energy Storage Technologies

open access: yesAdvanced Functional Materials, EarlyView.
Electrospinning transforms wood‐derived cellulose and lignin into architecturally defined, binder‐free carbon electrodes with tuneable porosity and functionality. This review shows how fibre design enables decoupled charge and mass transport, enhancing performance across battery systems, while identifying key challenges in spinnability, scalability ...
Michael W. Thielke   +3 more
wiley   +1 more source

Biodegradable and Biocompatible Functional Polymers for Biomedical Applications

open access: yesAdvanced Functional Materials, EarlyView.
Biodegradable and biocompatible functional polymers integrate electrical, mechanical, and stimuli‐responsive functionalities while enabling programmed degradation under physiological conditions. This review introduces recent advances in conductive, shape‐memory, self‐healing, photocurable, and adhesive polymer systems, emphasizing material design ...
Won Bae Han   +5 more
wiley   +1 more source

Parameter-Efficient Fine-Tuning for Foundation Models

open access: yes
This survey delves into the realm of Parameter-Efficient Fine-Tuning (PEFT) within the context of Foundation Models (FMs). PEFT, a cost-effective fine-tuning technique, minimizes parameters and computational complexity while striving for optimal ...
Wang, Yuandong   +5 more
core   +2 more sources

Efficient Low-Rank Adaptation for Sparse Large Language Model

open access: yesTsinghua Science and Technology
Existing Low-Rank Adaptation (LoRA) methods face challenges on sparse Large Language Models (LLMs) due to the inability to maintain sparsity. Recent works introduce methods that maintain sparsity by augmenting LoRA techniques with additional masking ...
Yuxuan Hu   +10 more
doaj   +1 more source

Two‐Photon Grayscale Lithography (2GL) of High Mechanical and Optical Quality 3D Silica Glass Microstructures

open access: yesAdvanced Functional Materials, EarlyView.
Two‐photon grayscale lithography (2GL) enables high‐speed and precise 3D printing of thermolyzed silica glass microstructures with optical‐grade surface quality, high quality factors, and mechanical strength utilizing a custom‐made pre‐glass resist based on polyhedral oligomeric silsesquioxane (POSS) modified with a high‐sensitivity Norrish type II ...
Jonathan L. G. Schneider   +4 more
wiley   +1 more source

Fine-tuning protein language models boosts predictions across diverse tasks

open access: yesNature Communications
Prediction methods inputting embeddings from protein language models have reached or even surpassed state-of-the-art performance on many protein prediction tasks.
Robert Schmirler   +2 more
doaj   +1 more source

Home - About - Disclaimer - Privacy