Chen Gao Signature

Tsinghua University · BNRist

Chen Gao

Explore the research
Chen Gao

About

Chen Gao is a Faculty Member (Research-track AP) of BNRist, Tsinghua University. He is a member of the EmbodiedAI Team at BNRist, a member of the Future Intelligence Lab, and a mentor at Zhongguancun Academy. He received his Ph.D. Degree (advised by Prof. Yong Li and Prof. Depeng Jin) and Bachelor's Degree from the Department of Electronic Engineering, Tsinghua University in 2021 and 2016, respectively.

His current research primarily focuses on world models, embodied intelligence, LLM/VLM agents, and multimodal reasoning systems, with 100+ CCF-A papers in top-tier venues and 10,000+ citations.

He was a visiting research scholar at the NExT Center, National University of Singapore, advised by Prof. Tat-Seng Chua and Prof. Xiangnan He. He is on the Stanford/Elsevier Top 2% Scientists List 2024 and 2025, received the ACL 2024 Outstanding Paper Award, and served as Area Chair for NeurIPS, ICML, ICLR, KDD, ACL ARR, WWW, and other venues.

Google Scholar Profile GitHub Projects
Research Focus

Research directions.

World Models

Benchmarks and generative models for embodied environments, long-horizon visual generation, and physical simulation.

Embodied Intelligence

Embodied agents that perceive, navigate, reason, and act in real-world city-scale spaces.

LLM/VLM Agents

Agentic reasoning, social simulation, recommendation, and decision support powered by foundation models.

Information Retrieval & Data Science

Early work in information retrieval and data mining, with continuing interests in recommendation, social computing, and data-driven intelligence.

目前团队尚有清华大学2027年入学的普博招生名额，欢迎在读硕士同学联系。 We welcome current Master's students to apply for a regular-track PhD position at Tsinghua University, starting in 2027.
Open Positions

Work with us

We welcome self-motivated interns, collaborators, Postdocs, PhD students, Master's students, and undergraduates interested in foundation models for embodied and open-world intelligence. If your interests match, please reach out with a concise CV, research interests, and representative work.

Opportunities & mentoring
Tsinghua undergraduates

Year-round research mentoring; we have experience guiding undergraduates across different years to publish high-quality papers.

External undergraduates / Master's students

Beijing/Shenzhen/Shanghai-based research internships (or remote) are welcome, with possible funding support and direct-admission or regular PhD application support for strong-fit candidates.

Early-stage PhD students

Joint mentoring and long-term collaboration are welcome, especially on world models and embodied agents.

Final-year PhD candidates

Postdoctoral positions are available for candidates who would like to continue independent research.

World Models Embodied Intelligence VLM/LLM Agents Multimodal Reasoning Drones and Robotics Computer Vision and Spatial Intelligence Social Simulation and Recommender Systems

Recent Highlights

New ResearchWorld Model

Worldscape-MoE

A unified mixture-of-experts world model for scalable heterogeneous action control.

Project Paper
IROS 2026Challenge

WorldArena 2.0 Challenge

The IROS 2026 WorldArena 2.0 Challenge is open for embodied world model evaluation. I lead this challenge, co-led with Dr. Yu Shang and Prof. Yong Li.

Challenge Homepage 机器之心报道
CVPR 2026Challenge

WorldArena 1.0 Challenge

The CVPR 2026 WorldArena Challenge evaluates embodied world models on perception quality and functional utility.

Challenge Homepage WorldArena
Spatial IntelligencePapers

Spatial Intelligence Papers

A focused collection of recent papers on urban embodied intelligence, 3D spatial reasoning, aerial navigation, and embodied world models.

Paper Collection Publications GitHub

News

  • Six papers on world models and spatial intelligence are accepted by NeurIPS 2026.
  • We release some preprint papers about world models and world action models, including TacWAM (Tactile World Action Model). Welcome to take a look!
  • The IROS 2026 WorldArena Challenge 2.0 competition opens. Please see the 机器之心报道 and challenge homepage for more details.
  • We released Worldscape-MoE, a unified mixture-of-experts world model for scalable heterogeneous action control. The paper is available on arXiv.
  • ANWM, our navigation world model paper, is accepted by ACM MM 2026.
  • The human-level spatial action benchmark paper is accepted by KDD 2026 as an Oral paper.
Earlier updates
  • We release some preprint papers about world models. Welcome to take a look!
  • Papers on world models are accepted by ICML 2026.
  • Papers on VLM and LLM agents are accepted by ACL 2026 main conference.
  • The CVPR 2026 WorldArena 1.0 Challenge is open, following the WorldArena benchmark and related media coverage.
  • Papers on world model and embodied agents are accepted by CVPR 2026 and ICME 2026.
  • We released WorldArena, a benchmark on embodied world models.
  • Papers are accepted by ACM TOIS and KDD 2026.
  • Four papers about visual language models and VLM agents are accepted by AAAI 2026.
  • The survey on large reasoning models is accepted by Cell Patterns.
  • Two papers are accepted by NeurIPS 2025. One is about world model and another is about token pruning for accelerating VLMs.
  • Four papers are accepted by EMNLP 2025 Main Conference: Embodied QA in City, theoretical analysis for LLM response, LLM-driven agents on social media, depression detection with LLM.
  • Three papers are accepted by ACM MM 2025: Embodied-R (Embodied spatial reasoning with RL post-training), AirScape (Aerial world model), and Open3D-VQA (3D VQA benchmark in open world).
  • We released some preprint papers about vision-language models: token pruning for accelerating VLMs, spatial reasoning of 3D VLMs, hallucinations of 3D VLMs.
  • We released some preprint papers on urban embodied intelligence: GeoNav, CityAVOS, DIMM.
  • We released some preprint papers about large language models empowered agents-based simulation: cultural bias, filter bubble, polarization.
  • One paper about dynamic graph neural networks is accepted by ACM TOIS.
  • Four papers on large language models are accepted by ACL 2025 and Findings of ACL 2025.
  • One survey about 3D LLMs is accepted by IJCAI 2025.
  • We released a survey on large language model agents for recommendation and search. Check the preprint paper for more information.
  • We released three benchmarks on outdoor embodied intelligence: Open3DVQA, CityEQA and UrbanVideo-Bench.
  • We released a preprint paper about evaluation of spatial intelligence.
  • We released a large-scale social simulator: AgentSociety 1.0. Check the preprint paper for more information.
  • One paper about large language models for session-based recommendation is accepted by ACM TOIS.
  • Two full papers and one resource-track paper are accepted by The Web Conference 2025.
  • We released a survey about reinforced reasoning with large language models.
  • We released a survey about responsible large language models.
  • Paper about graph neural network-based intelligent education is accepted by KDD 2025 (August track).
  • Two papers are accepted by AAAI 2025. One is about large language model and another is about learning from long behavioral sequence with spare attention.
  • Two papers about large language model-based recommender system are accepted by ACM TOIS (One is about collaborative filtering and another is about knowledge-based recommendation). We have also released a paper about large language model-based sequential recommendation.
  • Paper about complex network resilience prediction with graph neural network is published in Nature Communications. [Paper Link]
  • We released EmbodiedCity. For more details, please refer to the preprint paper, GitHub, and website. We will also organize a related workshop in ICLR 2025.
  • Paper about attacking large language models is accepted by NeurIPS 2024. [Paper Link]
  • I am selected as one of Stanford/Elsevier Top 2% Scientists.
  • Our survey paper on agent-based modeling and simulation with LLM agents is accepted by Humanities and Social Sciences Communications. For more details, please refer to the paper link.
  • We organized the Embodied Decision Intelligence Workshop.
  • I am selected as a Distinguished Reviewer of ACM TORS.
  • One paper about economic simulation with large language agents is accepted by ACL 2024 main conference. It receives ACL 2024 Outstanding Paper Award.
  • I am invited to give a talk about LLM-based combinational optimization in the 3rd forum on intelligent decision-making, Institute of Automation, CAS.
  • Our paper about user fatigue in recommender system is accepted by SIGIR 2024.
  • Our paper about stance detection with large language model agent is accepted by ICWSM 2024.
  • I am invited to give a talk about filter bubble in recommender system in Data Science Seminar of the School of EECS, The University of Queensland.
  • Three papers are accepted by ACM Web Conference 2024 (WWW 2024).
  • I am invited to give a talk about LLM-based social simulation and economic simulation in CCF RLChina 2023.
  • Our survey about causal inference-based recommender system is accepted by ACM Transactions on Information Systems (ACM TOIS). We have also released the 2nd Version on arXiv.
  • I am invited to give a talk about AutoML-based recommender system in CCF Student Pilot Program.
  • I am invited to give a talk about causal inference-based recommender system in DataFun Summit.
  • Two papers about recommendation are accepted by ACM International Conference on Web Search and Data Mining as oral paper (WSDM 2024).
  • One paper about urban knowledge graph is accepted by ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL 2023).
  • I will serve as Social Media Co-chair of SIGIR 2024.
  • One paper about short-video recommendation is accepted by ACM International Conference on Information and Knowledge Management (CIKM 2023).
  • One paper about recommendation robustness is accepted by ACM International Conference on Multimedia (ACMMM 2023).
  • One paper about short-video recommendation is accepted by ACM Conference on Recommender Systems (RecSys 2023).
  • I am invited to be a Program Committee (PC) Member of ACM International Conference on Information and Knowledge Management (CIKM 2023).
  • Three papers on recommendation, urban vulnerability, and living demands prediction are accepted by ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2023).
  • I am invited to be a Program Committee (PC) Member of ACM Multimedia Conference (ACMMM 2023).
  • One paper on knowledge graph-based recommendation is accepted by ACM Transactions on Information Systems (ACM TOIS).
  • One full paper and one short paper on micro-video recommendation and cold-start recommendation are accepted by ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2023).
  • I am invited to be a Program Committee (PC) Member of ACM Conference on Recommender Systems (RecSys 2023).
  • Three papers on controllable recommender system, sequential recommendation, and social recommendation are accepted by TheWebConf (TheWebConf/WWW 2023).
  • I am invited to be a Program Committee (PC) Member of ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2023).
  • One paper on disentangling geographical effect for point-of-interest recommendation is accepted by IEEE Transactions on Knowledge and Data Engineering (IEEE TKDE).
  • I am invited to be a Program Committee (PC) Member of International Conference on Artificial Intelligence and Statistics (AISTATS 2023).
  • We have released a survey on causal inference for recommender systems.
  • Our survey paper about graph neural network-based recommender systems is accepted by ACM Transactions on Recommender Systems.
  • One full research paper and one short research paper are accepted by ACM Conference on Information and Knowledge Management (CIKM 2022).
  • One paper on short-video recommendation is accepted by ACM Multimedia Conference (ACM MM 2022).
  • Two papers on user intention discovery and user decision-factor are accepted by ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2022).
  • One paper on disentangling social conformity and social influence for recommendation is accepted by IEEE Transactions on Knowledge and Data Engineering (IEEE TKDE).
  • Two short papers on sequential recommendation and session-based recommendation are accepted by ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2022).
  • We hosted the tutorial on "Graph Neural Networks for Recommender System" in ACM International Conference on Web Search and Data Mining (WSDM 2022).
  • I am invited to be a Program Committee (PC) Member of ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2022) and International Conference on Machine Learning (ICML 2022).
  • I am invited to give a talk about the GNN-RecSys survey in the Departmental Seminars (COMP 6911 & COMP 6912) Fall 2021 of the Dept. of ECE, HKUST.
  • Our tutorial proposal of "Graph Neural Networks for Recommender System" is accepted by ACM International Conference on Web Search and Data Mining (WSDM 2022).
  • We have released a survey entitled "Graph Neural Networks for Recommender Systems: Challenges, Methods, and Directions" at arXiv.
  • I am invited to be a Program Committee (PC) Member of TheWebConf 2022, International Conference on Learning Representations (ICLR 2022), ACM International Conference on Web Search and Data Mining (WSDM 2022), AAAI Conference on Artificial Intelligence (AAAI 2022), and International Joint Conference on Artificial Intelligence (IJCAI 2022).
  • I am invited to be a Program Committee (PC) Member of Conference on Neural Information Processing Systems (NeurIPS 2021).
  • I am invited to be a Program Committee (PC) Member of European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2021).

Survey

Workshop

Tutorial

Selected & published work

Publications

Google Scholar

Preprints

2026

Remembering What Matters: From Markovian to Subtask-Causal Memory in VLA Policies
P. Wang, F. Bai, Y. Li, T. Chou, Y. Xu, Y. Wen, Y. Yang, Chen Gao, Z. Zheng, F. Wu
Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026. (CCF-A)
DSPIGCN: Dual-stream Physics-informed Graph Convolutional Network for Reliable Pedestrian Trajectory Prediction
R. Guo, B. Chen, Z. Zhu, Chen Gao, Y. Zhao, Q. Yin
ACM Conference on Knowledge Discovery and Data Mining (KDD), 2026. (CCF-A)
DRAR: Diffusion-Based Relation Augmentation for Knowledge-aware Recommendation
Y. Peng, Chen Gao, T. Dan, Y. Li, X. Meng.
ACM Transactions on Information Systems (TOIS), 2026. (CCF-A)

2025

Social Bots Meet Large Language Model: Political Bias and Social Learning Inspired Mitigation Strategies
J. Piao, Z. Lu, Chen Gao, Y. Li.
ACM The Web Conference (TheWebConf/WWW), 2025. (CCF-A)
Breaking Student-Concept Sparsity Barrier for Cognitive Diagnosis
P. Shao, K. Zhang, Chen Gao, L. Chen, M. Cai, L. Wu, Y. Li, M. Wang.
Frontiers of Computer Science, 2025. (CCF-B)
Light Dynamic Graph Learning on Temporal Networks
Z. Pan, Chen Gao, F. Cai, H. Chen, Y. Li.
ACM Transactions on Information Systems (TOIS), 2025. (CCF-A)
Privacy-Preserving Recommendation with Coarse-Grained Spatiotemporal Contexts
L. Chen, Chen Gao, J. Lei, X. Du, X. Shi, H. Luo, D. Jin, Y. Li, M. Wang.
Science China Information Sciences, 2025. (CCF-A)

2024

2023

Modeling Multi-Grained User Preference in Location Visitation
Y. Qin, Chen Gao, T. Zhen, H. Wu, S. Wei, Y. Wang, L. Zhang, Y. Li.
ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (SIGSPATIAL, full, acceptance rate=15.7%), 2023. (CCF-C)

2022

2021

2020

2019

2018 & earlier

Academic Services

Conference Organizer

Conference Area Chair / SPC

Conference Reviewer / PC

Journal Reviewer

Awards

Research & collaboration

Let’s connect.

Interested in working together? I welcome research conversations, internships, and graduate or postdoctoral inquiries.

Email Chen Gaochgao96@tsinghua.edu.cn

A thoughtful first email

A short introduction is plenty. Please include:

  1. 01Your CV
  2. 02Research interests
  3. 03Representative work
  4. 04Expected timeline
Opportunities & research directions

World models · Embodied intelligence · VLM/LLM agents · Multimodal reasoning · Drones and robotics · Computer vision and spatial intelligence · Social simulation and recommendation.

  • Tsinghua undergraduatesYear-round mentoring toward high-quality research.
  • External undergraduates & Master’s studentsBeijing, Shenzhen, Shanghai or remote internships, with possible funding and PhD application support.
  • Early-stage PhD studentsJoint mentoring and long-term collaboration.
  • Final-year PhD candidatesPostdoctoral opportunities for independent research.

If you haven’t heard back after a week, a gentle follow-up is welcome. You can also reach me on Gmail with the same username.