Coarse-to-Fine Open-Set Graph Node Classification with Large Language Models

2025-12-18

AI Summary

Existing open-set classification methods for graph data treat out-of-distribution (OOD) samples monolithically as a single “rejection” class, lacking semantic granularity and interpretability. Method: We propose the first two-stage framework for semantic-level OOD classification on graphs: (1) coarse-grained OOD detection and interpretable semantic label generation via LLM-based prompt engineering; (2) joint GNN optimization for in-distribution (ID) node classification and OOD semantic classification, enhanced by multi-stage collaborative training and iterative label refinement. Crucially, our method operates without ground-truth OOD labels. Results: It achieves the first interpretable, fine-grained OOD semantic classification on graphs, improving OOD detection accuracy by 10% and attaining 70% accuracy on OOD semantic classification across mainstream graph benchmarks—substantially surpassing state-of-the-art. Our core contribution lies in breaking the conventional single-OOD-class assumption and establishing a unified modeling paradigm that jointly addresses ID recognition and OOD semantic understanding under graph-structured data.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Developing open-set classification methods capable of classifying in-distribution (ID) data while detecting out-of-distribution (OOD) samples is essential for deploying graph neural networks (GNNs) in open-world scenarios. Existing methods typically treat all OOD samples as a single class, despite real-world applications, especially high-stake settings such as fraud detection and medical diagnosis, demanding deeper insights into OOD samples, including their probable labels. This raises a critical question: can OOD detection be extended to OOD classification without true label information? To address this question, we propose a Coarse-to-Fine open-set Classification (CFC) framework that leverages large language models (LLMs) for graph datasets. CFC consists of three key components: a coarse classifier that uses LLM prompts for OOD detection and outlier label generation, a GNN-based fine classifier trained with OOD samples identified by the coarse classifier for enhanced OOD detection and ID classification, and refined OOD classification achieved through LLM prompts and post-processed OOD labels. Unlike methods that rely on synthetic or auxiliary OOD samples, CFC employs semantic OOD instances that are genuinely out-of-distribution based on their inherent meaning, improving interpretability and practical utility. Experimental results show that CFC improves OOD detection by ten percent over state-of-the-art methods on graph and text domains and achieves up to seventy percent accuracy in OOD classification on graph datasets.

Problem

• Classify in-distribution and detect out-of-distribution graph nodes
• Extend OOD detection to classification without true label information
• Leverage LLMs for interpretable OOD classification in graph datasets

Innovation

• Coarse-to-fine framework using LLMs for OOD detection
• GNN fine classifier trained with identified OOD samples
• LLM prompts generate semantic OOD labels for classification