Qianye Wu

Evaluating Modern Large Language Models on Low-Resource and Morphologically Rich Languages: A Cross-Lingual Benchmark Across Cantonese, Japanese, and Turkish

This work introduces a cross-lingual benchmark evaluating modern LLMs on low-resource and morphologically rich languages such as Cantonese, Japanese, and Turkish. It provides insights into multilingual generalization, morphology handling, and cultural adaptation of large language models across diverse linguistic structures.

Parallelism Meets Adaptiveness: Scalable Document Understanding in Multi-Agent LLM Systems

Accepted at: AAAI 2026 Workshop on Multi-Agent Path Finding (WoMAPF)

This paper proposes a parallel and adaptive multi-agent framework for document understanding. By coordinating LLM agents through a dynamic routing mechanism, it achieves scalable and efficient document parsing across long and complex texts.

AI-Driven Sentiment Analytics: Unlocking Business Value in the E-Commerce Landscape

This paper presents a sentiment analytics framework tailored for the e-commerce industry, integrating transformer-based models with domain-specific lexicons. The study demonstrates substantial business impact by enabling fast, secure, and reliable sentiment insights to support decision-making and enhance customer experience.