Haoyu Han

Efficient optimization and propagation for learning on web-scale graphs.

• LazyGNN: Large-Scale Graph Neural Networks via Lazy Propagation ICML ’23 Paper — LazyGNNCode — LazyGNN
• Alternately Optimized Graph Neural Networks ICML ’23 Paper — ALT-OPTCode — ALT-OPT

• Inverse APPNP: Solution for OGB-LSC 2022 MAG240M NeurIPS ’22 · Competition Paper — Inverse APPNPCode — Inverse APPNPResults — Inverse APPNP

Models that adapt to heterogeneous nodes, links, and graph spectra.

• Mixture of Link Predictors on Graphs NeurIPS ’24 Paper — Link-MoECode — Link-MoE
• Demystifying Structural Disparity in Graph Neural Networks: Can One Size Fit All? NeurIPS ’23 Paper — Structural DisparityCode — Structural Disparity

• Spectral-Aware Augmentation for Enhanced Graph Representation Learning CIKM ’24 Paper — GASSER
• Node-wise Filtering in Graph Neural Networks: A Mixture of Experts Approach Preprint · 2026 Paper — Node-MoE

Understanding and mitigating bias, collapse, and generalization failures in graph learning.

• Structural Fairness-aware Active Learning for Graph Neural Networks ICLR ’24 Paper — Structural FairnessCode — Structural Fairness
• Towards Label Position Bias in Graph Neural Networks NeurIPS ’23 Paper — Label Position BiasCode — Label Position Bias

• Unveiling Mode Connectivity in Graph Neural Networks KDD ’25 Paper — Mode Connectivity
• Enhancing Graph Representations Learning with Decorrelated Propagation KDD ’23 Paper — DePropCode — DeProp
• An Embarrassingly Simple Graph Heuristic Reveals Shortcut-Solvable Benchmarks for Sequential Recommendation NeurIPS ’26 · ED Paper — Benchmark Audit (TGH)Code — Benchmark Audit (TGH) Primarily Recommendation & E-Commerce

Explicit graph construction and exploration for reliable multi-step LLM reasoning.

• Reasoning by Exploration: A Unified Approach to Retrieval and Generation over Graphs WWW ’26 · Oral Paper — Reasoning by ExplorationCode — Reasoning by Exploration
• Reasoning with Graphs: Structuring Implicit Knowledge to Enhance LLMs Reasoning ACL Findings ’25 Paper — Reasoning with Graphs

• ConnectionMind: Leveraging Social Networks and Large Language Models for Personalized Recommendation at Meta CIKM ’26 · Applied Paper — ConnectionMind Primarily Recommendation & E-Commerce

Retrieval, filtering, integration, evaluation, and security over structured knowledge.

• RAG vs. GraphRAG: A Systematic Evaluation and Key Insights KDD ’26 · D&B Paper — RAG vs. GraphRAGCode — RAG vs. GraphRAG
• Retrieval-Augmented Generation with Graphs (GraphRAG) Preprint · 2025 Paper — GraphRAG SurveyResources — GraphRAG Survey

• Beyond Static Retrieval: Opportunities and Pitfalls of Iterative Retrieval in GraphRAG EMNLP ’26 · Main Paper — Beyond Static Retrieval
• Why Retrieval-Augmented Generation Fails: A Graph Perspective KDD ’26 Paper — Why RAG Fails
• Benchmarking Knowledge-Extraction Attack and Defense on Retrieval-Augmented Generation KDD ’26 · D&B · Oral Paper — RAG Attack & DefenseCode — RAG Attack & DefenseDataset — RAG Attack & Defense
• Empowering GraphRAG with Knowledge Filtering and Integration EMNLP ’25 Paper — GraphRAG-FI
• Mixture of Structural-and-Textual Retrieval over Text-rich Graph Knowledge Bases ACL Findings ’25 Paper — MoRCode — MoR
• From Flat to Structural: Enhancing Automated Short Answer Grading with GraphRAG EDM ’26 Paper — GraphRAG for Grading Primarily Education
• Enhancing LLM-Based Short Answer Grading with Retrieval-Augmented Generation EDM ’25 Paper — RAG for Grading Primarily Education

Language and video models as graph annotators, encoders, and predictors.

• Graph2Video: Leveraging Video Models to Model Dynamic Graph Evolution AAAI ’26 Paper — Graph2Video
• Label-free Node Classification on Graphs with Large Language Models (LLMs) ICLR ’24 Paper — LLM-GNNCode — LLM-GNNarchived

• GT2VEC: Large Language Models as Multi-Modal Encoders for Text and Graph-Structured Data KDD SKnow-LLM Workshop ’25 Paper — GT2VecCode — GT2Vec
• Understanding Graph Self-Supervised Pre-training under Distribution Shifts: A Scaling Law Perspective NeurIPS ’26 Paper — Graph SSL Scaling

Graph-based personalization, multimodal recommendation, and benchmark diagnostics.

• ConnectionMind: Leveraging Social Networks and Large Language Models for Personalized Recommendation at Meta CIKM ’26 · Applied Paper — ConnectionMind Also connected to: Structured LLM Reasoning
• An Embarrassingly Simple Graph Heuristic Reveals Shortcut-Solvable Benchmarks for Sequential Recommendation NeurIPS ’26 · ED Paper — Benchmark Audit (TGH)Code — Benchmark Audit (TGH) Also connected to: Reliability

• Enhancing ID and Text Fusion via Alternative Training in Session-based Recommendation IJCNLP–AACL ’25 Paper — AlterRecCode — AlterRec
• Amazon-M2: A Multilingual Multi-locale Shopping Session Dataset for Recommendation and Text Generation NeurIPS ’23 · D&B Paper — Amazon-M2Code — Amazon-M2Amazon Science — Amazon-M2Dataset — Amazon-M2
• STGCN: A Spatial-Temporal Aware Graph Learning Method for POI Recommendation ICDM ’20 Paper — STGCN

Knowledge-grounded assessment design and automated grading.

• From Flat to Structural: Enhancing Automated Short Answer Grading with GraphRAG EDM ’26 Paper — GraphRAG for Grading Also connected to: GraphRAG
• Enhancing LLM-Based Short Answer Grading with Retrieval-Augmented Generation EDM ’25 Paper — RAG for Grading Also connected to: GraphRAG

• Human-AI Collaboration for Knowledge-in-Use Assessment Design: Leveraging LLMs with RAG IEEE TALE ’25 Paper — Human–AI Assessment Design
• Constructing Educational Concept Maps with Multiple Relationships from Multi-source Data ICDM ’19 Paper — Educational Concept Maps

Graph learning for multimodal single-cell data integration.

• Multimodal Single-Cell Data Integration Project · NeurIPS ’21 Competition · 1st Competition page — DANCE Project repository — DANCE