Aount me

Hello! I’m Xinnan Dai, a third-year PhD student at Michigan State University, under the guidance of Prof. Jiliang Tang. Before that, I obtained bachelor’s and master’s degrees at Southwest University of Science and Technology and ShanghaiTech University, respectively. I was also a research intern at Microsoft Research Asia, mentored by Caihua Shan and Dongsheng Li.

My research interests include LLMs and graph reasoning, graph neural networks and AI for Science (specifically single-cell analysis).

I’m open to any internship opportunities.

contact: daixinna[at]msu[dot]edu

News

[Nov. 2025] The tutorial code for exploring implicit graph structures in decoder-only transformers is now available. Github (https://github.com/DDigimon/GraphGhost)

[Oct. 2025] Our Paper GraphGhost: Tracing Structures Behind Large Language Models is online.

[Oct. 2025] Our paper From Sequence to Structure: Uncovering Substructure Reasoning in Transformers and Uncovering Graph Reasoning in Decoder-only Transformers with Circuit Tracing got accepted by Neurips 2025. See you in San Diego! (Main Conference and Efficient Reasoning Workshop (Spotlight))

[Jun. 2025] Our paper PhysUniBench: An Undergraduate-Level Physics Reasoning Benchmark for Multimodal Models is online.

[May. 2025] Our paper Guided Structural Inference: Leveraging Priors with Soft Gating Mechanisms got accepted by ICML 2025.

[Feb. 2025] Our paper Exploring Graph Tasks with Pure LLMs: A Comprehensive Benchmark and Investigation is online.

[Feb. 2025] Our paper How Do Large Language Models Understand Graph Patterns? A Benchmark for Graph Pattern Comprehension got accepted by ICLR 2025.

[Aug. 2024] Our paper Revisiting the Graph Reasoning Ability of Large Language Models: Case Studies in Translation, Connectivity and Shortest Path is online.

[Aug. 2024] Our paper AI-powered omics-based drug pair discovery for pyroptosis therapy targeting triple-negative breast cancer got accepted by Nature Communications.

[Jan. 2024] We got Accelerate Foundation Models Academic Research Initiative (AFMR) funding from Microsoft Research.

Awards

Stars of Tomorrow Internship Program (MSRA)