Autonomous Data Science with LLM Agents
Autonomous Data Science with LLM Agents
I build autonomous data science agents that can perceive and reason over data, make effective modeling decisions, and operate reliably in real-world environments. My current work focuses on LLM agents for data science, particularly code generation and self-evolving agent systems.
Research Directions
Data-centric Agents
Perception, interpretation, adaptation, and governance mechanisms that give autonomous agents a reliable understanding of data.
Model-centric Agents
Reasoning, mathematical modeling, and inference-time methods that improve how agents formulate and solve data-science problems.
Systems-centric Agents
Execution, control, evaluation, and platform infrastructure for reliable long-horizon data-science agents.
A family of autonomous and self-evolving data-science agents for mathematical modeling, multimodal scientific discovery, and end-to-end data analysis.
A bilingual essay on moving from automated data-science workflows toward verifiable recursive self-improvement.
Selected Works
MM-Agent: LLM as Agents for Real-world Mathematical Modeling Problem
MM-Agent is an LLM agent framework for real-world mathematical modeling. It decomposes open-ended modeling into problem analysis, model formulation, computational solving, and report generation, enabling end-to-end solutions for real-world mathematical modeling tasks.
DSLighting
DSLighting is an LLM-driven autonomous data science execution engine that turns task descriptions and datasets into iterative code generation, execution, evaluation, and refinement workflows.
Recent Works
Expand / collapse list
(* Equal contribution)
- [EMNLP Findings] Fan Liu, Tianyu Pang, Chao Du, and Hao Liu. DSFlow: Evolutionary Workflow Optimization for Generalizable LLM-Based Data Science Automation. EMNLP 2026 Findings. [Findings]
- [Arxiv] Zherui Yang, Fan Liu, Hao Liu*. DSWorld: A Data Science World Model for Efficient Autonomous Agents. arXiv, 2026. [arXiv], [pdf], [Code]
- [KDD] Zherui Yang, Fan Liu, Yansong Ning and Hao Liu*. EvoDS: Self-Evolving Autonomous Data Science Agent with Capability Learning and Context Management. In Proceedings of the 32nd SIGKDD Conference on Knowledge Discovery and Data Mining, Jeju, South Korea, 2026. [arXiv], [Code] (CCF A)
- [ICLR] Fan Liu, Xiaozhao Zeng and Hao Liu. Towards Multimodal Data-Driven Scientific Discovery Powered by LLM Agents. In Proceedings of the Fourteenth International Conference on Learning Representations, Rio de Janeiro, Brazil, 2026. [OpenReview]
- [NeurIPS] Fan Liu, Jindong Han, Tengfei Lyu, Weijia Zhang, Zhe-Rui Yang, Lu Dai, Cancheng Liu, Hao Liu, Foundation Models for Scientific Discovery: From Paradigm Enhancement to Paradigm Transition, NeurIPS, 2025. [pdf], [Project] (CCF A) Position, Acceptance rate~6%
- [NeurIPS] Fan Liu*, Zherui Yang*, Cancheng Liu, Tianrui Song, Xiaofeng Gao, Hao Liu, MM-Agent: LLM as Agents for Real-world Mathematical Modeling Problem, NeurIPS, 2025. [OpenReview], [pdf], [Code], [Demo] (CCF A) 🔥🚀 Our MM-Agent system assists two undergraduate teams awarded F Award in 2025 MCM/ICM (top 2.0% among 27,456 human teams)
- [NeurIPS] Fan LIU, Wenshuo Chao, Naiqiang Tan, Hao Liu, Bag of Tricks for Inference-time Computation of LLM Reasoning, NeurIPS D&B, 2025. [OpenReview], [pdf], [Code] (CCF A)
- [WWW] Fan LIU, Hao Liu, Subgraph Federated Unlearning, WWW, 2025. [DOI], [OpenReview] (CCF A, Oral)
- [Arxiv] Fan LIU, Yue Feng, Zhao Xu, Lixin Su, Xinyu Ma, Dawei Yin, Hao Liu, JAILJUDGE: A Comprehensive Jailbreak Judge Benchmark with Multi-Agent Enhanced Explanation Evaluation Framework, Arxiv, 2024. [Project Page], [OpenReview], [pdf], [Code], [Dataset], [Model], [Coverage] 🔥🚀 Model 6000+ Downloads
- [NeurIPS] Zhao Xu, Fan LIU, Hao Liu, Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs, NeurIPS D&B, 2024. [pdf], [Code], [Coverage] (CCF A)
- [KDD] Fan LIU, Weijia Zhang, Hao Liu, Robust Spatiotemporal Traffic Forecasting with Reinforced Dynamic Adversarial Training, KDD, 2023. [arXiv] (CCF A)
- [NeurIPS] Fan LIU, Hao Liu, Wenzhao Jiang, Practical Adversarial Attacks on Spatiotemporal Traffic Forecasting Models, NeurIPS, 2022. [pdf], [Blog], [Code] (CCF A)
Education and Experience
Expand / collapse list
- 2022: Graduate student at HKUST(GZ)
- 2021: Intern at HKUST(GZ)
- 2020: Intern at MSRA (StarBridge Program)
- 2020: B.S. from UESTC
- 2019: Research visit at UBC
Awards, Acknowledgements, and Services
Expand / collapse list
- Reviewer for Conference: ICLR 2024-2025, NeurIPS 2023-2024, KDD 2023-2025, WWW 2025, AISTATS 2025, AdvML-Frontiers (ICML 2023 Workshop), FL4Data-Mining (KDD 2023 Workshop)
- Reviewer for Journal: ITS, Transactions On SMC: Systems, Physica A, TFS, TII
- TPC member: FL4Data-Mining (KDD 2023 Workshop)
- KDD Student Travel Award (2023)
- RBM Student Travel Grant (2023)
- Outstanding Undergraduate Thesis Award
- Outstanding Undergraduate Student
- Excellent Student Scholarship (2017-2020)