Researchers developed a parameter-efficient vision model tuning method to improve the automated screening of mild cognitive impairment from drawing tests.
A multi-agent simulation framework has been created to study political coalition formation while bypassing standard LLM neutrality biases.
CatalogAgent is a new supervisor-mediated system designed to improve the accuracy of product attribute extraction in e-commerce databases using generative AI.
A new psychometric scale has been developed to measure how undergraduate students rely on generative AI tools for academic writing tasks.
A new quantitative metric has been developed to measure the efficiency and utility of tool calls within LLM agent trajectories.
This paper introduces a multidimensional framework designed to standardize the evaluation of explainability methods in machine learning models.
A new framework aims to improve the quality and explainability of depression symptom annotations by aligning them with clinical diagnostic standards.
The proposed CoEvoT framework enhances graph-based reasoning in large language models by co-evolving chain-of-thought prompts to better handle distribution shifts.
ReportMedSAM utilizes natural language radiology reports to guide medical image segmentation, improving scalability and adaptability to new anatomical structures.
The study introduces a formal framework for managing multi-agent transactions between humans and their personal devices based on user intent.
This survey provides a unified mathematical taxonomy and reporting checklist for local additive feature attribution methods used in explainable AI.
This study compares the performance of generative AI against specialized supervised classification methods for the automated indexing of scientific literature.
The VLT foundation model integrates vision, language, and time-series data to improve industrial equipment monitoring and prognostic health management.
Researchers released a multimodal dataset combining visual and conversational data to improve emotion recognition and state monitoring for automotive safety systems.
The paper presents a flow-matching technique designed to incorporate specific constraints into robot motion generation tasks.
This paper proposes a reachability-aware pretraining method to improve the efficiency of reinforcement learning-based path exploration in temporal knowledge graph reasoning.
A new study highlights that current performance-optimization benchmarks for coding agents may be unreliable due to issues with runtime instability and inconsistent scoring metrics.
A new unified benchmark has been introduced to standardize the evaluation of foundation models applied to wireless communication tasks.
This paper introduces a gradient vector field surgery technique to address the over-confidence and miscalibration issues common in medical image segmentation models trained with standard loss functions.
A new method for generating adversarial text examples has been proposed to improve efficiency under low query budget constraints.
Meta has suspended an AI feature that allowed users to generate images using public Instagram photos following widespread criticism regarding its opt-out mechanism.
Apple has filed a lawsuit against OpenAI and Jony Ive's hardware startup, alleging a systematic theft of trade secrets by former Apple employees.
Apple is taking legal action against OpenAI, claiming the artificial intelligence company's leadership directed the theft of proprietary trade secrets.
China has completed its first 100,000-card GPU supercomputing cluster powered entirely by domestic hardware.