2022

Do Language Models Plagiarize?

Lee, Jooyoung, Le, Thai, Chen, Jinghui et al.

Understand

Past literature has illustrated that language models (LMs) often memorize parts of training instances and reproduce them in natural language generation (NLG) processes.

However, it is unclear to what extent LMs "reuse" a training corpus.
For instance, models can generate paraphrased sentences that are contextually similar to training samples.
In this work, therefore, we study three types of plagiarism (i.e., verbatim, paraphrase, and idea) among GPT-2 generated texts, in comparison to its training data, and further analyze the plagiarism patterns of fine-tuned LMs with domain-specific corpora which are extensively used in practice.

Built on

Karl Pearson and the chi-squared test
Robin L Plackett. 1983 · 1983
Earlier work this paper cites.
A learning algorithm for Boltzmann machines
David H Ackley, Geoffrey E Hinton, and Terrence J Sejnowski. 1985 · 1985
Earlier work this paper cites.
Okapi at TREC-3
Stephen E Robertson, Steve Walker, Susan Jones, Micheline M Hancock-Beaulieu, Mike Gatford, et al · 1995
Earlier work this paper cites.
Measures for corpus similarity and homogeneity. In Proceedings of the Third Conference on Empirical Methods for Natural Language Processing . 46–52
Adam Kilgarriff and Tony Rose. 1998 · 1998
Earlier work this paper cites.
In other (people’s) words: Plagiarism by university students–literature and lessons
Chris Park. 2003 · 2003
Earlier work this paper cites.
Plagiarism by academics: More complex than it seems
Roger Clarke. 2006 · 2006
Earlier work this paper cites.
Towards a definition of source-code plagiarism
Georgina Cosma and Mike Joy. 2008 · 2008
Earlier work this paper cites.
Academic writing and plagiarism: A linguistic analysis
Diane Pecorari. 2008 · 2008
Earlier work this paper cites.
Judging plagiarism: a problem of morality and convention
Julianne East. 2010 · 2010
Earlier work this paper cites.
Overview and comparison of plagiarism detection tools.. In Dateso . 161–172
Asim M El Tahir Ali, Hussam M Dahwa Abdulla, and Vaclav Snasel. 2011 · 2011
Earlier work this paper cites.
A Set-Based Approach to Plagiarism Detection.. In CLEF (Online Working Notes/Labs/Workshop)
Robin Küppers and Stefan Conrad. 2012 · 2012
Earlier work this paper cites.
Plagiarism meets paraphrasing: Insights for the next generation in automatic plagiarism detection
Alberto Barrón-Cedeño, Marta Vila, M Antònia Martí, and Paolo Rosso. 2013 · 2013
Earlier work this paper cites.
Using a Variety of n-Grams for the Detection of Different Kinds of Plagiarism
Prasha Shrestha and Thamar Solorio. 2013 · 2013
Earlier work this paper cites.
Adaptive algorithm for plagiarism detection: The best-performing approach at PAN 2014 text alignment competition. In International Conference of the Cross-Language Evaluation Forum for European Languages . Springer, 402–413
Miguel A Sanchez-Perez, Alexander Gelbukh, and Grigori Sidorov. 2015 · 2014
Earlier work this paper cites.
Arabic plagiarism detection using word correlation in N-Grams with K-overlapping approach. In Proceedings of the Workshops at the 7th Forum for Information Retrieval Evaluation (FIRE) . 123–125
Salha Alzahrani. 2015 · 2015
Earlier work this paper cites.
What is plagiarism and how to avoid it?
Ish Kumar Dhammi and Rehan Ul Haq. 2016 · 2016
Earlier work this paper cites.
Plagiarism detection in text documents using sentence bounded stop word n-grams
Deepa Gupta, K Vani, and LM Leema. 2016 · 2016
Earlier work this paper cites.
Exactus like: Plagiarism detection in scientific texts. In European conference on information retrieval . Springer, 837–840
Ilya Sochenkov, Denis Zubarev, Ilya Tikhomirov, Ivan Smirnov, Artem Shelmanov, Roman Suvorov, and Gennady Osipov. 2016 · 2016
Earlier work this paper cites.
Detection of idea plagiarism using syntax–semantic concept extractions with genetic algorithm
K Vani and Deepa Gupta. 2017 · 2017
Earlier work this paper cites.
A deep network model for paraphrase detection in short text messages
Basant Agarwal, Heri Ramampiaro, Helge Langseth, and Massimiliano Ruocco. 2018 · 2018
Earlier work this paper cites.

Similar

Hierarchical neural story generation
Original
Angela Fan, Mike Lewis, and Yann Dauphin. 2018 · 2018
Cited alongside, same era.
The secret sharer: Evaluating and testing unintended memorization in neural networks. In 28th USENIX Security Symposium (USENIX Security 19) . 267–284
Nicholas Carlini, Chang Liu, Úlfar Erlingsson, Jernej Kos, and Dawn Song. 2019 · 2019
Cited alongside, same era.
ArXiV Archive: A tidy and complete archive of metadata for papers on arxiv.org, 1993-2019
R. Stuart Geiger. 2019 · 2019
Cited alongside, same era.
The curious case of neural text degeneration
Original
Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. 2019 · 2019
Cited alongside, same era.
Automatic detection of generated text is easiest when humans are fooled
Original
Daphne Ippolito, Daniel Duckworth, Chris Callison-Burch, and Douglas Eck. 2019a · 2019
Cited alongside, same era.
Human and automatic detection of generated text
Daphne Ippolito, Daniel Duckworth, Chris Callison-Burch, and Douglas Eck. 2019b · 2019
Cited alongside, same era.
Roberta: A robustly optimized bert pretraining approach
Original
Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019 · 2019
Cited alongside, same era.
Leveraging BERT for extractive text summarization on lectures
Original
Derek Miller. 2019 · 2019
Cited alongside, same era.
Facebook FAIR’s WMT19 news translation task submission
Original
Nathan Ng, Kyra Yee, Alexei Baevski, Myle Ott, Michael Auli, and Sergey Edunov. 2019 · 2019
Cited alongside, same era.
Language models are unsupervised multitask learners
Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al · 2019
Cited alongside, same era.
Plagiarism detectors are a crutch, and a problem
Debora Weber-Wulff. 2019 · 2019
Cited alongside, same era.
Automatic plagiarism detection in obfuscated text
Alaa Saleh Altheneyan and Mohamed El Bachir Menai. 2020 · 2020
Cited alongside, same era.

Then

Exploring the limits of transfer learning with a unified text-to-text transformer
Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, Peter J Liu, et al · 2020
Later among the works it cites.
{ \{ Updates-Leak } \} : Data Set Inference and Reconstruction Attacks in Online Learning. In 29th USENIX Security Symposium (USENIX Security 20) . 1291–1308
Ahmed Salem, Apratim Bhattacharya, Michael Backes, Mario Fritz, and Yang Zhang. 2020 · 2020
Later among the works it cites.
Songmass: Automatic song writing with pre-training and alignment constraint
Original
Zhonghao Sheng, Kaitao Song, Xu Tan, Yi Ren, Wei Ye, Shikun Zhang, and Tao Qin. 2020 · 2020
Later among the works it cites.
Authorship attribution for neural text generation. In Conf. on Empirical Methods in Natural Language Processing (EMNLP)
Adaku Uchendu, Thai Le, Kai Shu, and Dongwon Lee. 2020 · 2020
Later among the works it cites.
Cord-19: The covid-19 open research dataset
Lucy Lu Wang, Kyle Lo, Yoganand Chandrasekhar, Russell Reas, Jiangjiang Yang, Darrin Eide, Kathryn Funk, Rodney Kinney, Ziyang Liu, William Merrill, et al · 2020
Later among the works it cites.
Attacking neural text detectors
Original
Max Wolff and Stuart Wolff. 2020 · 2020
Later among the works it cites.
Analyzing information leakage of updates to natural language models. In Proceedings of the 2020 ACM SIGSAC Conference on Computer and Communications Security . 363–375
Santiago Zanella-Béguelin, Lukas Wutschitz, Shruti Tople, Victor Rühle, Andrew Paverd, Olga Ohrimenko, Boris Köpf, and Marc Brockschmidt. 2020 · 2020
Later among the works it cites.
Detecting Fake News using Machine Learning: A Systematic Literature Review
Original
Alim Al Ayub Ahmed, Ayman Aljabouh, Praveen Kumar Donepudi, and Myung Suh Choi. 2021 · 2021
Later among the works it cites.
Extracting training data from large language models. In 30th USENIX Security Symposium (USENIX Security 21) . 2633–2650
Nicholas Carlini, Florian Tramer, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Ulfar Erlingsson, et al · 2021
Later among the works it cites.
Artificial text detection via examining the topology of attention maps
Original
Laida Kushnareva, Daniil Cherniavskii, Vladislav Mikhailov, Ekaterina Artemova, Serguei Barannikov, Alexander Bernstein, Irina Piontkovskaya, Dmitri Piontkovski, and Evgeny Burnaev. 2021 · 2021
Later among the works it cites.
Deduplicating training data makes language models better
Original
Katherine Lee, Daphne Ippolito, Andrew Nystrom, Chiyuan Zhang, Douglas Eck, Chris Callison-Burch, and Nicholas Carlini. 2021 · 2021
Later among the works it cites.
Investigating Memorization of Conspiracy Theories in Text Generation
Original
Sharon Levy, Michael Saxon, and William Yang Wang. 2021 · 2021
Later among the works it cites.
How much do language models copy from their training data? Evaluating linguistic novelty in text generation using RAVEN
Original
R Thomas McCoy, Paul Smolensky, Tal Linzen, Jianfeng Gao, and Asli Celikyilmaz. 2021 · 2021
Later among the works it cites.
Privacy regularization: Joint privacy-utility optimization in language models
Original
Fatemehsadat Mireshghallah, Huseyin A Inan, Marcello Hasegawa, Victor Rühle, Taylor Berg-Kirkpatrick, and Robert Sim. 2021 · 2021
Later among the works it cites.
TuringBench: A Benchmark Environment for Turing Test in the Age of Neural Text Generation. In Findings of Conf. on Empirical Methods in Natural Language Processing (EMNLP-Findings)
Adaku Uchendu, Zeyu Ma, Thai Le, Rui Zhang, and Dongwon Lee. 2021 · 2021
Later among the works it cites.
Counterfactual Memorization in Neural Language Models
Original
Chiyuan Zhang, Daphne Ippolito, Katherine Lee, Matthew Jagielski, Florian Tramèr, and Nicholas Carlini. 2021 · 2021
Later among the works it cites.
What Does it Mean for a Language Model to Preserve Privacy?
Original
Hannah Brown, Katherine Lee, Fatemehsadat Mireshghallah, Reza Shokri, and Florian Tramèr. 2022 · 2022
Closest in time.
Quantifying Memorization Across Neural Language Models
Original
Nicholas Carlini, Daphne Ippolito, Matthew Jagielski, Katherine Lee, Florian Tramer, and Chiyuan Zhang. 2022 · 2022
Closest in time.
Deduplicating training data mitigates privacy risks in language models
Original
Nikhil Kandpal, Eric Wallace, and Colin Raffel. 2022 · 2022
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

Reading the bibliography…