MemHunter: Automated and Verifiable Memorization Detection at Dataset-scale in LLMs

2024-12-10 · Zhenpeng Wu, Jian Lou, Zibin Zheng, Chuan Chen

Large language models (LLMs) have been shown to memorize and reproduce content from their training data, raising significant privacy concerns, especially with web-scale datasets. Existing methods for detecting memorization are primarily sample-specific, relying on manually crafted or discretely optimized memory-inducing prompts generated on a per-sample basis, which become impractical for dataset-level detection due to the prohibitive computational cost of iterating through all samples. In real-world scenarios, data owners may need to verify whether a susceptible LLM has memorized their dataset, particularly if the LLM may have collected the data from the web without authorization. To address this, we introduce MemHunter, which trains a memory-inducing LLM and employs hypothesis testing to efficiently detect memorization at the dataset level, without requiring sample-specific memory inducing. Experiments on models like Pythia and Llama demonstrate that MemHunter can extract up to 40% more training data than existing methods under constrained time resources and reduce search time by up to 80% when integrated as a plug-in. Crucially, MemHunter is the first method capable of dataset-level memorization detection, providing a critical tool for assessing privacy risks in LLMs powered by large-scale datasets.

📄 PDF Abstract BibTeX arXiv:2412.07261

Code (0)

등록된 구현이 없습니다.

Tasks

Memorization

Methods 이 논문이 사용한 방법론

LLaMA LLaMA is a collection of foundation language models ranging from 7B to 65B parameters. It is based on the transformer architecture with various improvements that were…

Pythia Pythia is a suite of decoder-only autoregressive language models all trained on public data seen in the exact same order and ranging in size from 70M to 12B parameters. The…

LiveSearchBench: An Automatically Constructed Benchmark for Retrieval and Reasoning over Dynamic Knowledge

2025-11-03 · Heng Zhou, Ao Yu, Yuchen Fan, Jianing Shi 외 arxiv

Evaluating large language models (LLMs) on question answering often relies on static benchmarks that reward memorization and understate the role of retrieval, failing to capture the dynamic nature of world knowledge. We …

Question Answering

Detecting Data Contamination in LLMs via In-Context Learning

2025-10-30 · Michał Zawalski, Meriem Boubdir, Klaudia Bałazy, Besmira Nushi 외 arxiv

We present Contamination Detection via Context (CoDeC), a practical and accurate method to detect and quantify training data contamination in large language models. CoDeC distinguishes between data memorized during train…

VeRA: Verified Reasoning Data Augmentation at Scale

2026-01-23 · Zerui Cheng, Jiashuo Liu, Chunjie Wu, Jianzhu Yao 외 arxiv

The main issue with most evaluation schemes today is their "static" nature: the same problems are reused repeatedly, allowing for memorization, format exploitation, and eventual saturation. To measure genuine AI progress…

Data Augmentation

LCMem: A Universal Model for Robust Image Memorization Detection

2025-12-16 · Mischa Dombrowski, Felix Nützel, Bernhard Kainz arxiv

Recent advances in generative image modeling have achieved visual realism sufficient to deceive human experts, yet their potential for privacy preserving data sharing remains insufficiently understood. A central obstacle…

Secret Stealing Attacks on Local LLM Fine-Tuning through Supply-Chain Model Code Backdoors

2026-04-30 · Zi Li, Tian Zhou, Wenze Li, Jingyu Hua 외 arxiv

Local fine-tuning datasets routinely contain sensitive secrets such as API keys, personal identifiers, and financial records. Although "local offline fine-tuning" is often viewed as a privacy boundary, we reveal that com…