Yansong Ning

Yansong is a first-year Ph.D. student in Artificial Intelligence Thrust at The Hong Kong University of Science and Technology, Guangzhou Campus. He continued his postgraduate studies after completing an M.Phil. degree at HKUST(GZ) under the supervision of Prof. Hao Liu and Prof. Hui Xiong. He interned at Didi Chuxing, working with L Lab. Prior to that, he obtained his Bachelor's degree in the School of Data Science at TYUT. During undergraduate studies, he also interned at Guangzhou HKUST FYTRI, collaborating with Alibaba Cloud Intelligence Group.

His recent research focuses on reasoning LLMs and agent. Specifically, his research goal is to understand and improve the "agentic reasoning capability" of LLMs (automatically plan, execute external tools, and reflect on tool responses to explore, verify, and refine intermediate actions in multi-step reasoning process), guaranteeing their practical applicability in real-world deployment, e.g., optimizing agent to satisfy user's objectives via multiturn interaction.

What's
New

[Sep 2025] Pleasure to release DeepTravel, an end-to-end agentic reinforcement learning framework for building autonomous travel agent!!
[Jul 2025] Two papers were accepted to ECML-PKDD and ECAI, about Federated Graph Learning and Federated Fine-tuning! Congrates to my co-author!
[May 2025] Thrilled to release Long⊗Short, an efficient reasoning framework that enables two LLMs to collaboratively solve the problem. Feel free to use our released OpenMath-ThoughtChunk1.8K on Hugging Face for thought-level reasoning analysis!!
[May 2025] DiMA is accepted to KDD 2025!!
 [April 2025] Pleasure to give a talk about LLM-based UrbanKG Construction in The Geographical Society of China 2025 Academic Annual Meeting at Zhejiang University!!
[Feb 2025] Thrilled to release DiMA, an LLM-powered ride-hailing assistant deployed on DiDi Chuxing. Thanks for the guidance and support from L Lab!!
[Sep 2024] UrbanKGent is accepted to NeurIPS 2024!
[Dec 2023] Pleasure to join DiDi Chuxing as a research intern, collaborating with the L Lab on the LLM pre-training, post-training and LLM-based ride-hailing assistant deployment.
[Sep 2023] UUKG is accepted to NeurIPS 2023!
[Aug 2022] Pleasure to start the summer internship at the Guangzhou HKUST FYTRI, collaborating with Alibaba Cloud Intelligence Group on the projected related to urban knowledge graph.

Featured Research Projects

Description of image

DiMA: An LLM-Powered Ride-Hailing Assistant at DiDi

KDD 2025
Yansong Ning, Shuowei Cai, Wei Li, Jun Fang, Naiqiang Tan, Hua Chai, Hao Liu

[Paper][Code] Full Deployment on DiDi Chuxing

[KDD 2025] Yansong Ning, Shuowei Cai, Wei Li, Jun Fang, Naiqiang Tan, Hua Chai, Hao Liu. DiMA: An LLM-Powered Ride-Hailing Assistant at DiDi. [Paper] [Code] Full Deployment on DiDi Chuxing
Description of image

DeepTravel: An End-to-End Agentic Reinforcement Learning Framework for Autonomous Travel Planning Agents

Preprint
Yansong Ning, Rui Liu, Jun Wang, Kai Chen, Wei Li, Jun Fang, Kan Zheng, Naiqiang Tan, Hao Liu

[Paper] Beta Deployment on DiDi Chuxing

[Preprint] Yansong Ning, Rui Liu, Jun Wang, Kai Chen, Wei Li, Jun Fang, Kan Zheng, Naiqiang Tan, Hao Liu. DeepTravel: An End-to-End Agentic Reinforcement Learning Framework for Autonomous Travel Planning Agents. [Paper] Beta Deployment on DiDi Chuxing
Description of image

UrbanKGent: A Unified Large Language Model Agent Framework for Urban Knowledge Graph Construction

NeurIPS 2024
Yansong Ning, Hao Liu

[Paper] [Code] [Slides] [Video]

[NeurIPS 2024] Yansong Ning, Hao Liu. UrbanKGent: A Unified Large Language Model Agent Framework for Urban Knowledge Graph Construction. [Paper] [Code] [Slides] [Video]
Description of image

UUKG: Unified Urban Knowledge Graph Dataset for Urban Spatiotemporal Prediction

NeurIPS 2023
Yansong Ning, Hao Liu, Hao Wang, Zhenyu Zeng, Hui Xiong

[Paper] [Code] [Slides] [Video]

[NeurIPS 2023] Yansong Ning, Hao Liu, Hao Wang, Zhenyu Zeng, Hui Xiong. UUKG: Unified Urban Knowledge Graph Dataset for Urban Spatiotemporal Prediction. [Paper] [Code] [Slides] [Video]
Description of image

Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Multi-Turn Reinforcement Learning

Preprint
Yansong Ning, Wei Li, Jun Fang, Naiqiang Tan, Hao Liu

[Paper] [Code] [Data]

[Preprint] Yansong Ning, Wei Li, Jun Fang, Naiqiang Tan, Hao Liu. Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Multi-Turn Reinforcement Learning. [Paper] [Code] [Data]
Description of image

USTBench: Benchmarking and Dissecting Spatiotemporal Reasoning of LLMs as Urban Agents

Preprint
Siqi Lai, Yansong Ning, Zirui Yuan, Zhixi Chen, Hao Liu

[Paper] [Code]

[Preprint] Siqi Lai, Yansong Ning, Zirui Yuan, Zhixi Chen, Hao Liu. USTBench: Benchmarking and Dissecting Spatiotemporal Reasoning of LLMs as Urban Agents. [Paper] [Code]

Publications And Manuscripts

Publications
Preprints

Education

  • Sep 2025 - Present, Guangzhou, The Hong Kong University of Science and Technology
    Doctor of Philosophy in Artificial Intelligence
    Supervisor: Prof. Hao Liu
  • Sep 2023 - Jun 2025, Guangzhou, The Hong Kong University of Science and Technology
    Master of Philosophy in Artificial Intelligence
    Supervisor: Prof. Hao Liu & Prof. Hui Xiong
  • Sep 2019 - Jun 2023, Taiyuan, Taiyuan University of Technology
    Bachelor of Data Science and Technology of Big Data
    Supervisor: Prof. Li Wang

Work Experience

Teaching/Conference Services

Teaching Assistant: AIAA 3111: Introduction to Data Mining (Spring, 2025)

Session Chair: KDD'2025

Reviewer: NeurIPS'2024'2025, ICLR'2025'2026, ICML'2025, KDD'2026

Awards

  • RBM Travel Grant, 2024
  • Outstanding Undergraduate Thesis, 2023 (Top 0.2%, 10 out of 7762)
  • Finalist Winner of the Mathematical Contest in Modeling (MCM), 2022 (Top 2%)
  • Second Prize in the Contemporary Undergraduate Mathematical Contest in Modeling (CUMCM), 2020 (Top 2%)
Flag Counter