Junjie Wu

Reward Modeling & Agent Evaluation

Working under the GenAI org. Deployed an unified MLLM-as-a-Judge framework with enhanced post-training techniques for advertisement generation.

Working on evaluating the inductive reasoning ability of large language models.

NLP Center Worked on adversarial attack and defense for neural machine translation.

Advised by Prof. Minlie Huang (PI of the CoAI Lab) Worked on tracking and controlling topic transition in document-grounded dialog system.

Advised by Prof. David Jurgens (PI of the Blablablab) Worked on predicting prosocial (defined by many metrics like healthy, supportive, politeness) outcomes in online conversations from large scale of Reddit data.

Advised by Prof. Yongfeng Huang (PI of the NGN-Lab) Worked on designing a novel multi-task learning based framework to generate high-quality sentence representations for emotion classification using labeled-text and emotion lexicon.

Advised by Prof. Dit-Yan Yeung Thesis title: Exploring the Intelligence of Natural Language Processing Systems and Large Language Models: Perspectives on Learning and Understanding