Citation Recommendation
MasterSet: A Benchmark for Must-Cite Citation Recommendation
A large-scale benchmark for identifying papers that researchers must cite, built from 150,000+ AI/ML papers across 15 leading venues.
About Me
I am an Assistant Professor of Computer Science at Northern Illinois University (NIU). I received my Ph.D. in Computer Science from Virginia Tech in December 2023, where I was advised by Prof. Chang-Tien Lu at the Sanghani Center for Artificial Intelligence and Data Analytics.
I study how machine learning reasons over structure, from graph learning and neural algorithmic reasoning to combinatorial and mixed-integer nonlinear optimization and graph-based retrieval-augmented generation. I also work on AI for science: retrieval over the scientific literature, such as finding the papers a researcher must cite, and machine learning for planetary science. With my students I apply these ideas to software vulnerability detection and to testing how far large language models can be trusted. Feel free to reach out if you see potential alignment and are interested in collaborating.
At NIU, I co-founded the NIU AI and organize the AI & ML research seminar series. NIU students can subscribe here to participate.
News
Hiring, Funding & Committee Policy (Read before emailing)
I am hiring undergraduate research assistants and PhD students to work on AI and Machine Learning.
Notice to New/Current NIU Graduate Students:
- I do not provide financial support or RA positions to students I have no prior academic connection with.
- Please do not cold email me for funding if you haven't taken my courses (and earned an 'A') or regularly participated in my research seminars.
- Unsolicited emails regarding funding will be ignored.
Thesis/Dissertation Committee Requests:
Students who would like me to serve on their committee are expected to participate in my research seminar during the semesters they are working with me. Regular participation allows me to follow your progress and give meaningful input at your proposal and defense.
- Oct. 2026 Two papers are accepted by IEEE BigData 2026.
- Oct. 2026 A high school student I mentored over the summer had a paper accepted at the IEEE BigData 2026 High School Symposium.
- Oct. 2026 Serving as Digital Infrastructure Chair for JCDL 2026.
- Jul. 2026 Two papers are accepted by SIAM SDM.
- Jul. 2026 Gave an AI workshop for the Huskies BELONG Career and College Prep Program at NIU.
- Jun. 2026 One paper was presented at the EarthVision Workshop at CVPR 2026.
- May 2026 Received an AI Curricular Innovation Grant from the Center for Innovative Teaching and Learning (CITL) at NIU.
- Apr. 2026 One paper is accepted by JGR: Machine Learning and Computation.
- Jan. 2026 Launched a six-week Secure-AI Project at Marmion Academy, in collaboration with Dr. Cansu Tatar, empowering high school students’ artificial intelligence and cybersecurity literacy.
Earlier news
- Dec. 2025 Served as Digital Infrastructure Chair for JCDL 2025.
- Nov. 2025 Served as Poster Chair for IEEE ICDM 2025.
- Jul. 2025 One paper is accepted by ACM SIGSPATIAL 2025.
- May 2025 One paper is accepted by KDD 2025.
- May 2025 Chaired a session at SIAM SDM 25’.
- May 2025 Hosted a tutorial session at SIAM SDM 25’.
- Apr. 2025 One paper is accepted by Knowledge and Information Systems (KAIS).
- Oct. 2024 Hosted a tutorial session at SIAM MDS 24’.
- Oct. 2024 Hosted a tutorial session at CIKM 24’.
- Jun. 2024 Our tutorial on Unifying Spectral and Spatial Graph Neural Network at CVPR 24’ is live!
- Aug. 2024 Joined the Department of Computer Science at Northern Illinois University.
Research
Neural Combinatorial Optimization
Network Interdiction Goes Neural
Solving Network Interdiction Problems with Graph Neural Networks
Graph Neural Network
Infinitely Deep Graph Transformation Networks
Deep Equilibrium Model for Node-Edge Co-Evolution Graph Neural Network
Paper Lists
Curated, annotated reading lists on GitHub, each organized around one question about its field and refreshed from my paper collection.
- Machine Learning for Graph Drawing: neural networks that generate, update, select or optimize the layout of node-link diagrams.
- Neural Algorithmic Reasoning and Combinatorial Optimization: neural networks that execute classical algorithms or solve combinatorial problems, read as one literature.
- Scientific Literature Retrieval: citation recommendation, academic search, citation-informed document embeddings and reranking.
- AI-Based Vulnerability Detection: papers, datasets and benchmarks on learning-based source-code vulnerability detection.
Recent Publications
(2026).
CiteRoute: Facet-Routed Retrieval for Core Citation Recommendation.
In IEEE BigData ‘26, to appear.
(2026).
From Causal Plausibility to Causal Reliability: Evaluating LLMs as Calibrated Direct Causal-Edge Classifiers.
In IEEE BigData ‘26 (short paper), to appear.
(2026).
Delta-Graph Signatures for Post-Edit Risk Screening in Knowledge Editing.
In SDM ‘26.
(2026).
MasterSet: A Large-Scale Benchmark for Must-Cite Citation Recommendation in the AI/ML Literature.
In SDM ‘26.
(2026).
CraterBench-R: Instance-Level Crater Retrieval for Planetary Scale.
In CVPR Workshops ‘26 (EarthVision).
Blog
Your PhD Application, Seen from the Other Side
A few years ago I was the applicant. Now I read the applications. Here is what actually carries information for me, what carries none, and what quietly works against you.
Research Without Black Boxes
Intellectual independence and a refusal to leave important things as black boxes: two things I hope my students carry into their research.
When a Language Model Refers to Itself, What Is "Itself"?
A Turing-complete Transformer inherits Kleene-style self-reference, but the self it refers to is the prompt, not the weights. Notes on what that does and does not say about recursive self-improvement.
Three Fixed Points Hiding in "Recursive Self-Improvement"
Kleene’s recursion theorem, the Y combinator, and a contraction mapping all give you x = f(x). They answer different questions, and it helps to keep them apart before talking about self-improving AI.
Treat Me Like an Agent
The students who get the most out of me are doing something that looks a lot like prompt engineering — whether or not they would describe it that way.