Matrix-Driven Instant Review: Confident Detection and Reconstruction of LLM Plagiarism on PC
It addresses intellectual property concerns for LLM developers by offering an efficient and accessible detection method, though it appears incremental as it builds on existing detection approaches with specific improvements.
The paper tackles the problem of detecting plagiarism in large language models (LLMs) by proposing Matrix-Driven Instant Review (MDIR), which accurately reconstructs weight correspondences and provides rigorous p-value estimation, achieving reliable detection even after extensive transformations like random permutations and continual pretraining with trillions of tokens, all within an hour on a single PC.
In recent years, concerns about intellectual property (IP) in large language models (LLMs) have grown significantly. Plagiarizing other LLMs (through direct weight copying, upcycling, pruning, or continual pretraining) and claiming authorship without properly attributing to the original license, is a serious misconduct that can lead to significant financial and reputational harm to the original developers. However, existing methods for detecting LLM plagiarism fall short in key areas. They fail to accurately reconstruct weight correspondences, lack the ability to compute statistical significance measures such as $p$-values, and may mistakenly flag models trained on similar data as being related. To address these limitations, we propose Matrix-Driven Instant Review (MDIR), a novel method that leverages matrix analysis and Large Deviation Theory. MDIR achieves accurate reconstruction of weight relationships, provides rigorous $p$-value estimation, and focuses exclusively on weight similarity without requiring full model inference. Experimental results demonstrate that MDIR reliably detects plagiarism even after extensive transformations, such as random permutations and continual pretraining with trillions of tokens. Moreover, all detections can be performed on a single PC within an hour, making MDIR both efficient and accessible.