Team Poznan at SemEval-2026 Task 13: Detecting Machine-Generated Code with Multiple Programming Languages, Generators, and Application Scenarios

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy, Software Development & Engineering · Depth: Expert, quick

Summary

Team Poznan at SemEval-2026 Task 13 introduced a novel method for detecting machine-generated code, a critical task for software security and quality. Addressing the limitations of traditional stylistic or statistical approaches, which advanced code generation models increasingly circumvent, their solution leverages Graph Neural Networks (GNNs). This GNN-based classifier represents code as a program dependency graph to capture its structural characteristics. The team demonstrated that their approach significantly outperforms both traditional and embedding-based methods on benchmark datasets, achieving improved accuracy and robustness in identifying code generated by various techniques. This work underscores the potential of GNNs for a deeper, structural understanding of code authorship.

Key takeaway

For AI Security Engineers or Machine Learning Engineers focused on code integrity, this research suggests a critical shift. If you are currently relying on stylistic or embedding-based methods for detecting machine-generated code, you should explore Graph Neural Networks (GNNs). Implementing GNNs, which analyze code's structural characteristics via program dependency graphs, can significantly improve detection accuracy and robustness against advanced code generation models, enhancing software supply chain security.

Key insights

GNNs effectively detect machine-generated code by analyzing structural characteristics via program dependency graphs.

Principles

• Structural code analysis improves detection.
• GNNs surpass traditional code authorship methods.
• Program dependency graphs reveal code origins.

Method

The method involves representing code as a program dependency graph and then applying a Graph Neural Network (GNN) classifier. This GNN analyzes structural characteristics to identify machine-generated code.

In practice

• Implement GNNs for code authorship detection.
• Use program dependency graphs for code analysis.
• Evaluate GNNs against embedding-based methods.

Topics

• Machine-Generated Code Detection
• Graph Neural Networks
• Program Dependency Graphs
• Code Authorship
• Software Security
• SemEval-2026 Task 13

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, AI Security Engineer

Related on AIssential

• The Code Whisperer: LLM and Graph-Based AI for Smell and Vulnerability Resolution — A hybrid AI framework unifies code smell and vulnerability resolution using graph analysis and LLMs.
• HybridCodeAuthorship: A Benchmark Dataset for Line-Level Code Authorship Detection — The increasing hybrid nature of codebases necessitates fine-grained, line-level AI authorship detection benchmarks.
• Foundation Models for Structured Data — Applying transformer-style attention to databases viewed as graphs enables foundation models for structured data, overcoming manual feature engineering.
• Unlocking the relational power of graph neural networks in fraud prevention — Graph Neural Networks enhance fraud detection by modeling relational data, uncovering networked fraud invisible to traditional tabular methods.
• Graphical-Probabilistic Modeling of Generative Flows in LLM-Native Software Systems — Generation Networks provide a principled, graphical probabilistic framework for designing and analyzing LLM-native software systems.
• Structure and Relationships: Graph Neural Networks and a Pytorch Implementation — GNNs model interconnected data by transforming, aggregating, and updating node features based on neighbor relationships.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.