mdok of KInIT: Robustly Fine-tuned LLM for Binary and Multiclass AI-Generated Text Detection

TL;DR

This paper presents mdok, a robust fine-tuning approach for smaller LLMs to detect AI-generated texts, achieving top performance in binary and multiclass classification tasks, even with out-of-distribution data.

Contribution

The paper introduces mdok, a novel fine-tuning method for smaller LLMs that enhances robustness and accuracy in detecting AI-generated texts across multiple classification scenarios.

Findings

Achieved 1st rank in Voight-Kampff AI detection challenge

Demonstrated robustness to out-of-distribution data

Effective in both binary and multiclass detection tasks

Abstract

The large language models (LLMs) are able to generate high-quality texts in multiple languages. Such texts are often not recognizable by humans as generated, and therefore present a potential of LLMs for misuse (e.g., plagiarism, spams, disinformation spreading). An automated detection is able to assist humans to indicate the machine-generated texts; however, its robustness to out-of-distribution data is still challenging. This notebook describes our mdok approach in robust detection, based on fine-tuning smaller LLMs for text classification. It is applied to both subtasks of Voight-Kampff Generative AI Detection 2025, providing remarkable performance (1st rank) in both, the binary detection as well as the multiclass classification of various cases of human-AI collaboration.