Benyou Wang

AceGPT enhances Arabic language processing tailored for local culture and values.

Research reveals significant biases in human and LLM evaluations of responses.

New benchmarks reveal challenges for MLLMs in real-world tasks with long contexts.

Assessing how LLMs adapt to new information and biases.

A new benchmark assesses LLMs' abilities in mathematical modeling processes.

Training open-source LLMs enhances optimization modeling for industry applications.

LLMs are assisting doctors, improving patient care while reducing risks of misinformation.

LongLLaVA improves multi-image understanding for various applications.

TRIM method reduces image tokens in multi-modal language models while maintaining performance.

A new framework identifies when multimodal models use inappropriate training data.