Kai-Wei Chang

Examining biases in image generation and their societal impacts.

EMO-SUPERB project enhances speech emotion recognition through improved techniques and community collaboration.

A new system to evaluate audio codec performance across various applications.

This study analyzes language biases affecting gender and race in various texts.

This study shows how context improves machine understanding of images and questions.

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

LLM-A* merges traditional algorithms with language models for efficient path planning.

A new system enhances accuracy and reliability in text generation from RALMs.

Enhancing user engagement in large vision-language models through proactive communication.

A method to generate keyphrases based on user needs for better content summarization.

This study explores the trade-off between diversity and factual accuracy in AI-generated images.

A new method bridges language models and knowledge graphs for better information access.

New methods improve automated question-answer relationships in writing analysis.

JourneyBench tests AI models on unusual and fictional images.

A new method helps computers identify objects using fewer images and simple language.

A safety database to respect cultural norms and legal guidelines globally.