Latest Articles for Model Optimization

A new method enhances the alignment of language models using multiple references.

New layer pruning technique enhances model efficiency and accuracy.

A new method enhances the fine-tuning of large language models for better efficiency.

This paper discusses Online Ensembles of Basis Expansions to improve machine learning.

FedMR tackles challenges in federated learning with partial class data, enhancing model performance.

ETHER introduces a cost-effective way to fine-tune large machine learning models.

A new method improves efficient deep learning models through exact orthogonality.

New methods enhance main task performance using auxiliary data without extra computation costs.

This article examines layer normalization's role in improving neural network classification.

A new framework improves pruning methods for large language models without retraining.

Examining the saturation effect in Kernel Ridge Regression and its implications for predictions.

VTrans method significantly reduces transformer model sizes without sacrificing performance.

Study reveals effective techniques to enhance multimodal large language models.

New adaptable models can meet diverse needs without retraining.

A framework to enhance Gaussian Process Regression's predictions and uncertainty measures.

New methods improve machine learning models across diverse environments.

Research outlines techniques to improve efficiency in serving LoRA adapters.

SHiRA improves model switching efficiency in AI without losing key concepts.

PruningBench offers a standardized way to evaluate pruning methods, enhancing model efficiency in machine learning.

Examining the unusual attention behavior in Transformer models.

Model merging combines different AI models for improved performance across tasks.

Discover how genetic algorithms can refine hyperparameter tuning in machine learning models.

A new framework enhances large model performance efficiently during fine-tuning.

CPT improves black-box model performance without direct access to internal parameters.

M IST enhances interaction between visual and language models for better performance.

Learn how step size affects gradient descent in logistic regression.

A new method improves model accuracy and efficiency in fluctuating data environments.

ISQuant offers a new approach to quantization for efficient model deployment.

Discover how adaptive dynamic quantization enhances VQ-VAE models for better data representation.

A method to enhance model efficiency in machine learning through effective pruning strategies.

New framework improves efficiency of Vision Transformers while maintaining accuracy.

A novel method enhances image classification using topological data analysis and knowledge distillation.

New methods improve continual learning and adaptability of large pre-trained models.

A new method to enhance pre-trained models using selective fine-tuning.

A flexible model architecture that enhances Transformer efficiency and performance.

New methods reduce memory usage while maintaining performance in LLMs.

A new method to select data augmentations improves model performance on time series tasks.

Introducing a new method to enhance efficiency in large language models through pruning.

Examining dynamic methods for optimizing machine learning model training.

LeanQuant improves model size and quality through advanced quantization techniques.