Hongru WANG Hongru WANG

Hongru WANG

Ph.D. Candidate

Department of Systems Engineering and Engineering Management,

The Chinese University of Hong Kong

About Me CV Research Statement Teaching Statement

I am currently a final-year Ph.D. Candidate at The Chinese University of Hong Kong (expected to graduate on July 2025 ), very honored and fortunate to be supervised by Prof. Kam-Fai WONG. Before that, I obtained my Bachelor's degree and Master's degree from the Communication University of China and The Chinese University of Hong Kong in 2019 and 2020, respectively. During my PhD, I am happy and honored to visit The University of Edinburgh (EdinburghNLP) and University of Illinois Urbana-Champaign (BlenderLab), working closely with Prof. Jeff Z. Pan, Prof. Pasquale Minervini and Prof. Heng Ji. Besides that, I am co-founder and organizer of NLP Academic Exchange Platform (NICE), which provide a platform to share and discuss recent progress in NLP. 📚📚📚

My research focus revolves around reasoning and acting of personalized language agents, designed to seamlessly unifing them from tool perspective such as regarding reasoning as internal cognitive tools while acting as external physical tools instead of treat them in isolation. My long-term objective is to achieve the ''impossible triangle'' between safety, personalization and autonomy of language agent. If you are interested in my research (more details can be found below), I dedicate 60 minutes per week to chat with persons from all over the world, feel free to say hi if you want to discuss or cooperate. Schedule meeting with me! 📆

Research Interest

  • Dialogue System: Prior to LLMs, I focus on task-oriented dialogue system, especially the natural langugae understanding (MCML), and dialogue policy learning (PPO-Off-Comb, Survey of DPL). We build a first Cantonese task-oriented dialogue dataset -- KddRES. Then I turn to open-domain dialogue system, especially the interal reasoning capabilities of LLMs (Cue-CoT) and how they interect with external multiple knowledge sources (SAFARI, UniMS-RAG). We recently survey the evolution of dialogue system based on language model (from ToD to ODD, and then unified dialogue system) (LM-based DS).

  • Tool Learning: Besides only modelling physical tools, such as models and APIs, I am interested in how internal cognitive tools (TPE), like different conversational strategies and reasoning methods, can be combined with external physical tools. I am attracted to the whole life cycle of tool learning, such as tool creation (UniRetriever), tool selection (Self-DC, AppBench) and tool utilization (M3SUM). Utilizing tool learning to empower LLMs for real-world interaction is both intriguing and impactful. You can check our latest tutorial at SIGIR: Empoweing LLMs: Tool learning for Real-world Interaction.

  • Language Agents: More broadly, I believe dialogue will be the entrance of next human-ai interaction, similar like human beings. To this end, I also focus other aspects of diague agents, including but not limited to safety (INDust, Self-Guard), role-playing (REGA), memory (PerLTQA), factuality (K-Dial) and proactivity (Pro-CoT).

News

Competition and Conference [Full]

  • May 2025: Nine papers are accepted by ACL 2025: 3 Main and 6 Findings, including 2 first-author papers and 3 (co)first-author papers.
  • Apr 2025: We are so exited to introduce OTC and ToolRL. We believe OTC will be the foundation of agentic RL like the ReAct of Agent.
  • Jan 2025: Three papers are accepted by NAACL 2025, including one first author work: Self-DC that empower language agent when to rely on internal knowledge and when to call external tools. 😍

Preprint

Draft and Under Review

OTC: Optimal Tool Calls via Reinforcement Learning
Hongru Wang, Cheng Qian, Wanjun Zhong, Xiusi Chen, et al., Mengdi Wang, Kam-Fai Wong, Heng Ji
Arxiv (First work to optimize tool-use behaviors of Agent)
ToolRL: Reward is All Tool Learning Needs
Cheng Qian, et al., Hongru Wang, Xiusi Chen, Dilek Hakkani-Tür, Gokhan Tur, Heng Ji
Arxiv (First comprehensive analysis of reward design for tool learning)
RM-R1: Reward Modeling as Reasoning
Xiusi Chen, Gaotang Li, et al., Hongru Wang, et al.,, Tong Zhang, Hanghang Tong, Heng Ji
Arxiv (First work to cast reward modeling as reasoning)
UniMS-RAG: A Unified Multi-source RAG for Personalized Dialogue Systems
Hongru Wang, Wenyu Huang, Yang Deng, et. al, Jeff Z Pan, Kam-Fai Wong
Under Review
A Survey of the Evolution of Language Model-Based Dialogue Systems
Hongru Wang, Lingzhi Wang, Yiming Du, Liang Chen, Jingyan Zhou, Yufei Wang, Kam-Fai Wong
Under Review

Selected Publications

Journal and conference [Full]

Self-Reasoning Language Models: Unfold Hidden Reasoning Chains with Few Reasoning Catalyst
Hongru WANG, Deng Cai, Wanjun Zhong, Shijue Huang, Jeff Z. Pan, Zeming Liu, Kam-Fai Wong
Reasoning and Planning for Large Language Models of ICLR 2025
ACL 2025, Findings (data synthesise to self-improve itself for LLMs)
Rethinking Stateful Tool Use in Multi-Turn Dialogues: Benchmarks and Challenges
Hongru WANG, Wenyu Huang, et al., Zeming Liu, Jeff Z. Pan, Kam-Fai Wong
ACL 2025, Findings
Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering
YuZhao, et al., Hongru WANG, Xuanli He, Kam-Fai Wong, Pasquale Minervini
NAACL 2025 (tool conflicts with each other, how to detect and control the behaviors)
Oral = 246 / 3185 ~= 7.7%
Self-DC: When to Reason and When to Act? Self Divide-and-Conquer for ..
Hongru Wang, Boyang Xue, Baohang Zhou, et. al, Kam-fai Wong
NAACL 2025 (self-aware knowledge boundary to manage tool calls)
Oral = 246 / 3185 ~= 7.7%
AutoPSV: Automated Process-Supervised Verifier
Jianqiao Lu, Zhiyang Dou, Hongru WANG, Zeyu Cao, Jianbo Dai, Yingjia Wan, Yinya Huang, Zhijiang Guo
NeurIPS 2024
AppBench: Planning of Multiple APIs from Various APPs for Complex Instruction
Hongru WANG, Rui Wang, Boyang XUE, Heming Xia, Jingtao Cao, Zeming Liu, Jeff Z. Pan, Kam-Fai Wong
EMNLP 2024 (apple intelligence, graph execution, permission management) Code Poster
Empowering Large Language Models: Tool Learning for Real-World Interaction
Hongru WANG, Yujia Qin, Yankai Lin, Jeff Z. Pan and Kam-Fai Wong
Tutorial, SIGIR 2024 (first comprehensive tutorial about tool learning)
Large Language Models as Source Planner for Personalized ..
Hongru Wang, Minda Hu, Yang Deng, et. al, Irwin King, Kam-Fai Wong
EMNLP 2023, Findings
Best Paper Award @ International Doctoral Forum (2023) Code Poster PPT
Cue-CoT: Chain-of-thought Prompting for Responding to In-depth Dialogue ..
Hongru Wang*, Rui Wang*, Fei Mi, Yang Deng, Zezhong Wang, Bin Liang, Ruifeng Xu, Kam-Fai Wong
EMNLP 2023, Findings Code Poster PPT

Projects

Hongru WANG, Gabriel, Kam-Fai Wong (2020). Dialogue Collection Platform for Restaurant in HK. We build the website to mimic the coversation between consumer and restaurant and collect the data to build task-oriented dialogue system. Website


Hongru WANG, Gabriel, Changjian Liu, Kam-Fai Wong (2020). SMP 2020 ECDT Few-shot Spoken Language Understanding. We use ERNIE, ERNIE+CRF, ERNIE+BiLSTM+CRF to solve this problem, and different trick like mask strategy. We get Top 3 at the leaderboard. Leaderboard


Hongru WANG, Li Min (2019). A Game about Error Correction for Cantonese. Using js-cookie, javascript, bootstracp DEMO

Experience

Intern and Volunteer

2024.08 - 2025.05

ByteDance Seed-LLM-Harizon (Doubao)
Research Intern

2020.08 - 2021.09

CUHK MoE Key Lab of High Confidence Software Technologies
Research Assistant

2019.04 - 2019.07

Kuaishou Music and Video Recommendation
Data Intern

Awards & Grants

Competitions, Scholarship

  • Reaching Out Award (2023-2024)
  • Overseas Research Attachment Programme (ORAP 2023-2024)
  • Technology and Business Development Fund (TBF22ENG004)
  • Top 1 at Online Safety Price organized by AI Singapore (WWW 2024)
  • Top 10 at ICLR 2021 Workshop MLPCP Track 1 (rank 7th)
  • Third Price at SMP2020-ECDT Few-shot Spoken Language Understanding
  • Distinguished Academic Performance Scholarship (2019-2020), CUHK CSE
  • Top 10 at SemEval2020-Task4: CommonSense Detection and Explanations
  • Meritorious Winner, Mathematical Contest In Modeling (2018)
  • A software copyright of “WeCampus WeChat mini-program” (2018SR562540)
  • Third Prize of Internet + College Students Innovation and Entrepreneurship Competition China National Radio Scholarship

Services

Reviewers, Workshops

TA

Teaching Assistant, Student Helper

  • SEEM 3450 Engineering Innovation and Entrepreneurship
  • SEEM 3490 Information Systems Management
  • SEEM 5730 / ECLT 5910 Information Technology Management

Tools

Open Source Tools, just for fun

Contact

  • You can find me in wechat public account: 地球漫游必备手册 -> 慧用智能
  • hrwang@se.cuhk.edu.hk
  • Hong Kong, ShenZhen, Beijing, GuangZhou, HangZhou, Singapore, Bali