Junjie Wu 伍君杰

👋 Hi!

I am a Senior Researcher on Tencent’s Hunyuan Foundation Model Team, working on models such as Hy4 and Hy4 preview. My work focuses on agentic evaluation (agent behavior analysis and next-generation evaluation pipelines) and efficient benchmarking, with the goal of turning evaluation into actionable feedback for agent development while reducing its cost.

I received my Ph.D. in Artificial Intelligence from HKUST in January 2026, advised by Prof. Dit-Yan Yeung.

Previously, I was a Research Scientist Intern with Meta’s GenAI team and a visiting Ph.D. student at Yale University, working with Prof. Arman Cohan. I have also collaborated closely with Mo Yu and Lemao Liu at Tencent AI Lab and WeChat AI.

Junjie Wu

News

Education / Experience

  1. – Present

    Tencent Hunyuan

    Senior Researcher

    1. Hy4
    2. Hy3
  2. –

    Hong Kong University of Science and Technology

    Ph.D. in Artificial Intelligence · Advised by Prof. Dit-Yan Yeung

    1. –

      GenAI @ Meta

      Research Scientist Intern

    2. –

      Yale University

      Visiting Ph.D. Student · Yale NLP

    3. –

      WeChat AI @ Tencent

      Research Intern

    4. –

      Tencent AI Lab

      Research Intern · NLP Group

  3. –

    Sun Yat-sen University

    B.S. in Statistics · GPA: 3.8/4.0

    1. –

      University of Michigan

      Visiting Student · Ann Arbor

Research

My current research focuses on agentic evaluation (agent behavior analysis and next-generation evaluation pipelines) and efficient benchmarking.

My previous work explored how to make large language and vision-language models more intelligent and reliable, with an emphasis on:

Publications (* Equal contribution)

Preprints
Beyond Outcomes: Dual-View Relational Learning for Efficient Agent Benchmarking
Xinshuai Guo, Junjie Wu, Dolly Deng, Yinghui Li, Hai-Tao Zheng, Suncong Zheng, and Maxm Pan
[Paper]
Bi-Level Prompt Optimization for Multimodal LLM-as-a-Judge
Bo Pan, Xuan Kan, Kaitai Zhang, Yan Yan, Shunwen Tan, Zihao He, Zixin Ding, Junjie Wu, and Liang Zhao
[Paper]
Accepted
Mindscape-Aware Retrieval Augmented Generation for Improved Long Context Understanding
Yuqing Li, Jiangnan Li, Zheng Lin, Ziyan Zhou, Junjie Wu, Weiping Wang, Jie Zhou, and Mo Yu
EMNLP 2026
[Paper] [Model]
Situated Embedding Models for Context-Aware Dense Retrieval
Junjie Wu, Jiangnan Li, Yuqing Li, Lemao Liu, Liyan Xu, Jiwei Li, Dit-Yan Yeung, Jie Zhou, and Mo Yu
ACL 2026
[Paper] [Model]
Multi-Task Reinforcement Learning for Enhanced Multimodal LLM-as-a-Judge
Junjie Wu, Xuan Kan, Zihao He, Shunwen Tan, Bo Pan, and Kaitai Zhang
ACL 2026 Industry Track
[Paper]
Ref-Long: Benchmarking the Long-context Referencing Capability of Long-context Language Models
Junjie Wu*, Gefei Gu*, Yanan Zheng, Dit-Yan Yeung, and Arman Cohan
ACL 2025
[Paper] [Project] [Code]
Understanding LLMs' Fluid Intelligence Deficiency: An Analysis of the ARC Task
Junjie Wu, Mo Yu, Lemao Liu, Dit-Yan Yeung, and Jie Zhou
NAACL 2025 (Oral)
[Paper] [Project] [Code]
The Stochastic Parrot on LLM’s Shoulder: A Summative Assessment of Physical Concept Understanding
Mo Yu*, Lemao Liu*, Junjie Wu*, Tsz Ting Chung*, Shunchi Zhang*, Jiangnan Li, Dit-Yan Yeung, and Jie Zhou
NAACL 2025 (Oral)
[Paper] [Project] [Code]
Unified Triplet-Level Hallucination Evaluation for Large Vision-Language Models
Junjie Wu*, Tsz Ting Chung*, Kai Chen*, and Dit-Yan Yeung
TMLR 2025
[Paper] [Project] [Code]
Rethinking Targeted Adversarial Attacks for Neural Machine Translation
Junjie Wu, Lemao Liu, Wei Bi, and Dit-Yan Yeung
ICASSP 2024
[Paper] [Code]
Towards General Error Diagnosis via Behavioral Testing in Machine Translation
Junjie Wu, Lemao Liu, and Dit-Yan Yeung
EMNLP 2023 Findings (Presented at the GenBench Workshop @EMNLP 2023)
[Paper] [Code]
Conversations Gone Alright: Quantifying and Predicting Prosocial Outcomes in Online Conversations
Jiajun Bao*, Junjie Wu*, Yiming Zhang*, Eshwar Chandrasekharan, and David Jurgens
The Web Conference 2021
[Paper] [Code]