Latest Articles for Natural Language Processing

A novel method improves understanding of language model outputs.

A method to rewrite texts while protecting individuals' privacy.

A new approach improves dialogue systems by combining topic and rhetorical structures.

Research shows diverse instructions improve language model performance in unseen tasks.

New method increases text generation speed using adaptive candidate selection.

This study compares LLMs integrated with various symbolic solvers for logical reasoning tasks.

This study enhances relation extraction by linking it with natural language inference techniques.

Introducing the Robust Partial Wasserstein distance for better data comparison.

CLIP exhibits strength in handling data imbalance in visual and language tasks.

Mamba-2 combines SSMs and Transformers for improved efficiency in language tasks.

New methods aim to ensure safe and high-quality text generation from language models.

A new method improves reward models using synthetic critiques for better alignment.

A new method improves image generation accuracy from text prompts.

Small wording changes can greatly influence language model outcomes.

A new method enhances relation extraction across long documents.

New methods enhance language model outputs while maintaining grammar rules.

A new method enhances vision-language models' performance with known and unknown classes.

A new approach to enhance small language models using sparse activation techniques.

Manticore automates the creation of hybrid language models, improving efficiency and performance.

Combining visual-language models with reinforcement learning improves task completion efficiency.

Exploring safety, reliability, and ethical issues in language models.

Researchers develop a model to improve graph learning using data from various sources.

New methods improve language model predictions under varying input conditions.

TransCLIP enhances predictions by integrating visual and textual data in Vision-Language Models.

A framework to enhance neural networks by integrating human knowledge into learning algorithms.

SpatialRGPT enhances object arrangement understanding in Vision Language Models.

A method using MCMC for effective negative sample generation in contrastive learning.

A new system improves serving large language models across various GPU configurations.

A new method enhances self-training for language agents using reflection models.

Examining how recurrent models can approximate functions based on prompts.

A novel method enhances language models for better efficiency and performance.

Integrating SysCaps into energy modeling simplifies decision-making and improves predictions.

This paper presents a new approach to enhance KGQA performance using GNNs and LLMs.

This research reveals how images and text interact in reasoning tasks.

Examining over-optimization in DAAs and its impact on model performance.

Innovative approach to guide large language models using self-assessment.

The Block Transformer improves text processing speed and efficiency in language models.

A look at uncertainty types and their importance in language models.

A look at models that operate without matrix multiplication for better efficiency.

A new method improves translation quality through effective data augmentation.