Fine-Tuning Pre-Trained Code Models for AI-Generated Code Detection
This work addresses the practical need for detecting AI-generated code, but the results are incremental and competitive rather than breakthrough.
The paper presents a system for detecting AI-generated code, achieving macro-F1 scores of 0.737 on binary classification (6th out of 81 teams) and 0.422 on 11-class attribution (7th out of 34 teams) in the SemEval-2026 shared task.
This paper describes the system submitted by team \textbf{Archaeology} to SemEval-2026 Task~13 on AI-generated code detection. The shared task consists of three subtasks; we participate in Subtask-A (binary classification: human-written vs.\ AI-generated code) and Subtask-B (11-class attribution of the generating model). Starting from a TF-IDF and Logistic Regression baseline, we fine-tune four pre-trained code models (CodeBERT, GraphCodeBERT, UniXcoder, and CodeT5+) with separate strategies for each subtask. For Subtask-A, we use leave-one-language-out cross-validation, code augmentation, chunked inference with trimmed-mean aggregation, and threshold calibration on a difficult dataset. For Subtask-B, we use sandwich token packing, class-balanced loss, and multi-seed ensembling with test-time augmentation. Our best submissions obtain macro-F1 scores of 0.737 on Subtask-A (6th/81 teams) and 0.422 on Subtask-B (7th/34 teams).