Pan Li
Associate Professor
Studies machine learning on graphs, including geometric deep learning for science and graph methods for trustworthy LLMs.
Pan Li's lab at Georgia Tech works on trustworthy & safe ML, graph learning and geometry processing & modeling. 60 papers since 2023, 41 at top venues, mostly ICLR and NeurIPS. 7 PhD students. 20 alumni, most often to Google and Amazon. Applications close .
At a glance
- papers, 2023–2026
- 16/ year
- at T1 venues
- 68%
- PhD students
- 7
- MS students
- 0
- alumni listed
- 20
Papers by yearfrom the lab homepage and DBLP, 2023–2026
- 13 papers2026 · T1 8 · collecting
- 17 papers2025 · T1 14
- 21 papers2024 · T1 13
- 9 papers2023 · T1 6
T1 venuesOtherVenues: ICLR, NeurIPS, ICML, WWW, Findings of EMNLP
Research focuspaper titles that use each topic's words
- Graph learning33 papers
- geometric deep4 papers
- llms2 papers
- Trustworthy & safe ML1 paper
- Geometry processing & modeling0 papers
Paper list
From the lab homepage and DBLP.
202613 papers
- ICLR (T1)∇-Reasoner: LLM Reasoning via Test-Time Gradient Descent in Latent Space (opens in a new tab)
- ACM Trans. Knowl. Discov. DataA Survey of Deep Graph Learning under Distribution Shifts: From Graph Out-of-Distribution Generalization to Adaptation (opens in a new tab)
- TPAMI (T1)How Can State Space Models Enhance Machine Learning on Graphs? (opens in a new tab)
- ICLR (T1)Measuring Physical-World Privacy Awareness of Large Language Models: An Evaluation Benchmark (opens in a new tab)
- ICLR (T1)MoEEdit: Efficient and Routing-Stable Knowledge Editing for Mixture-of-Experts LLMs (opens in a new tab)
- —Powers of Magnetic Graph Matrix: Fourier Spectrum, Walk Compression, and Applications (opens in a new tab)
- ICLR (T1)Reducing Belief Deviation in Reinforcement Learning for Active Reasoning of LLM Agents (opens in a new tab)
- ICLR (T1)Scalable Spatio-Temporal SE(3) Diffusion for Long-Horizon Protein Dynamics (opens in a new tab)
- —The Trojan Knowledge: Bypassing Commercial LLM Guardrails via Harmless Prompt Weaving and Adaptive Tree Search (opens in a new tab)
- ISQEDThor: Towards General Directed Circuit Graph Encoder with Sample-Efficient Graph Contrastive Learning (opens in a new tab)
- WWW (T1)Towards A Universal Graph Structural Encoder (opens in a new tab)
- —Towards Direct Latent-Space Synthesis for Parallel Branches in LLM-Agent Workflows (opens in a new tab)
202517 papers
- ICLR (T1)Convergent Privacy Loss of Noisy-SGD without Convexity and Smoothness (opens in a new tab)
- NeurIPS (T1)Differentially Private Relational Learning with Entity-level Privacy Guarantees (opens in a new tab)
- Evaluating Unlearning with Knowledge Correlation and Confidence Awareness (code) NeurIPSDo LLMs Really Forget? Evaluating Unlearning with Knowledge Correlation and Confidence Awareness (opens in a new tab)
- ICML (T1)Generalization Principles for Inference over Text-Attributed Graphs with Large Language Models (opens in a new tab)
- NeurIPS (T1)Graph-KV: Breaking Sequence via Injecting Structural Biases into Large Language Models (opens in a new tab)
- JMLR (T1)Improving Graph Neural Networks on Multi-node Tasks with the Labeling Trick (opens in a new tab)
- ICLR (T1)LayerDAG: A Layerwise Autoregressive Diffusion Model for Directed Acyclic Graph Generation (opens in a new tab)
- ICML (T1)Rethinking Addressing in Language Models via Contextualized Equivariant Positional Encoding (opens in a new tab)
- NeurIPS (T1)RoFt-Mol: Benchmarking Robust Fine-Tuning with Molecular Graph Foundation Models (opens in a new tab)
- ICLR (T1)Simple is effective: The roles of graphs and large language models in knowledge-graph-based retrieval-augmented generation (opens in a new tab)
- NeurIPS (T1)Time-o1: Time-Series Forecasting Needs Transformed Label Alignment (opens in a new tab)
- Findings of EMNLPTowards Universal Debiasing for Language Models-based Tabular Data Generation (opens in a new tab)
202421 papers
- J. Appl. Comput. Topol.Ambiguities in neural-network-based hyperedge prediction (opens in a new tab)
- NeurIPS (T1)Certified Machine Unlearning via Noisy Stochastic Gradient Descent (opens in a new tab)
- NeurIPS (T1)Differentially Private Graph Diffusion with Applications in Personalized PageRanks (opens in a new tab)
- Trans. Mach. Learn. Res.DIG-MILP: a Deep Instance Generator for Mixed-Integer Linear Programming with Feasibility Guarantee (opens in a new tab)
- TKDE (T1)Distance Information Improves Heterogeneous Graph Neural Networks (opens in a new tab)
- —Exploring the Opportunities and Challenges of Graph Neural Networks in Electrical Engineering (opens in a new tab)
- WSDM (T1)GAD-NR: Graph Anomaly Detection via Neighborhood Reconstruction (opens in a new tab)
- —GeSS: Benchmarking Geometric Deep Learning under Scientific Applications with Distribution Shift (opens in a new tab)
- NeurIPS (T1)GeSS: Benchmarking Geometric Deep Learning under Scientific Applications with Distribution Shifts (opens in a new tab)
- ICML (T1)Graph As Point Set (opens in a new tab)
- Trans. Mach. Learn. Res.GraphMaker: Can Diffusion Models Generate Large Attributed Graphs? (opens in a new tab)
- NeurIPS (T1)Langevin Unlearning: A New Perspective of Noisy Gradient Descent for Machine Unlearning (opens in a new tab)
20239 papers
- NeurIPS (T1)Differentially Private Decoupled Graph Convolutions for Multigranular Topology Protection (opens in a new tab)
- ICLR (T1)Equivariant Hypergraph Diffusion Neural Operators (opens in a new tab)
- WWW (T1)Expressive and Efficient Representation Learning for Ranking Links in Temporal Graphs (opens in a new tab)
- ICCV (T1)Extensible and Efficient Proxy for Neural Architecture Search (opens in a new tab)
- (codes)Interpretable Geometric Deep Learning via Learnable Randomness Injection (opens in a new tab)
- —Semi-supervised Graph Neural Network for Particle-level Noise Removal (opens in a new tab)
- (codes)Structural Re-weighting Improves Graph Domain Adaptation (opens in a new tab)
- VLDB (T1)SUREL+: Moving from Walks to Sets for Scalable Subgraph-based Graph Representation Learning (opens in a new tab)
- ICLR (T1)Unsupervised Learning for Combinatorial Optimization Needs Meta Learning (opens in a new tab)
Where alumni went20 listed, 71% to industry
Industry 5Academia 2
Most often to
- Amazon
- Microsoft
Apply to Georgia TechFall 2027
Georgia Tech Computer Science PhD
- Round
- Single round
- Opens
- —
- GRE
- optional
- English
- TOEFL/IELTS
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- USUCSD
Yoav Freund
- Learning theory & optimization
- USMIT
Stephen Bates
- Trustworthy & safe ML
- USMichigan
Michael Wellman
- Reasoning & neuro-symbolic AI
- CNTsinghua
Keyou You
- Reinforcement learning
- USCMU
Zachary C. Lipton
- Trustworthy & safe ML
- USUIUC
Maxim Raginsky
- Learning theory & optimization
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage and DBLP
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
Georgia Techcloses