This paper describes our system for SemEval-2026 Task 13 Subtask B, which requires attributing source code to either a human author or one of 10 LLM families. Guided by dataset analysis, we identify three practical challenges: formatting fingerprints discarded by tokenizers, heterogeneous code lengths, and extreme class imbalance. We build on unixcoder-base with Explicit Stylistic Prompting, Language-Aware Truncation, and imbalance-aware training (Focal Loss, GeM pooling, multi-sample dropout, and bucket batching). Our system achieves 0.434 Macro F1 on the official hidden test set, ranking 4th out of 34 teams with only 125M parameters. Controlled 5-fold cross-validation confirms that each component contributes to the final system, and a formatting-normalization study quantifies the model’s reliance on formatting cues.

In this paper, we propose a novel framework for the tutor identification track of the BEA 2025 shared task (Track 5). Our framework integrates data-algorithm co-design, dynamic class compensation, and structured prediction optimization. Specifically, our approach employs noise augmentation, a fine-tuned DeBERTa-v3-small model with inverse-frequency weighted loss, and Hungarian algorithm-based label assignment to address key challenges, such as severe class imbalance and variable-length dialogue complexity. Our method achieved 0.969 Macro-F1 score on the official test set, securing second place in this competition. Ablation studies revealed significant improvements: a 9.4% gain in robustness from data augmentation, a 5.3% boost in minority-class recall thanks to the weighted loss, and a 2.1% increase in Macro-F1 score through Hungarian optimization. This work advances the field of educational AI by providing a solution for tutor identification, with implications for quality control in LLM-assisted learning environments.