KInIT at SemEval-2024 Task 8: Fine-tuned LLMs for Multilingual Machine-Generated Text Detection
Michal Spiegel, Dominik Macko
TL;DR
The paper tackles multilingual machine-generated text detection under Subtask A of SemEval-2024 Task 8. It presents an ensemble approach that fine-tunes two 7B-LMs (Falcon-7B and Mistral-7B) with parameter-efficient fine-tuning and fuses their predictions with three zero-shot statistical detectors via a two-step majority vote, plus per-language thresholds, using language IDs. The main finding is that this setup achieves competitive accuracy and strong generalization, ranking fourth and approaching the winner's performance; post-deadline experiments with late-stage refinements show substantial gains in AUC ROC. The work highlights practical benefits of combining fine-tuned LLMs with statistical signals for multilingual detection in resource-constrained settings.
Abstract
SemEval-2024 Task 8 is focused on multigenerator, multidomain, and multilingual black-box machine-generated text detection. Such a detection is important for preventing a potential misuse of large language models (LLMs), the newest of which are very capable in generating multilingual human-like texts. We have coped with this task in multiple ways, utilizing language identification and parameter-efficient fine-tuning of smaller LLMs for text classification. We have further used the per-language classification-threshold calibration to uniquely combine fine-tuned models predictions with statistical detection metrics to improve generalization of the system detection performance. Our submitted method achieved competitive results, ranking at the fourth place, just under 1 percentage point behind the winner.