DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature

Eric Mitchell, Yoonho Lee, Alexander Khazatsky, Christopher D Manning, Chelsea Finn
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:24950-24962, 2023.

Abstract

The increasing fluency and widespread usage of large language models (LLMs) highlight the desirability of corresponding tools aiding detection of LLM-generated text. In this paper, we identify a property of the structure of an LLM’s probability function that is useful for such detection. Specifically, we demonstrate that text sampled from an LLM tends to occupy negative curvature regions of the model’s log probability function. Leveraging this observation, we then define a new curvature-based criterion for judging if a passage is generated from a given LLM. This approach, which we call DetectGPT, does not require training a separate classifier, collecting a dataset of real or generated passages, or explicitly watermarking generated text. It uses only log probabilities computed by the model of interest and random perturbations of the passage from another generic pre-trained language model (e.g., T5). We find DetectGPT is more discriminative than existing zero-shot methods for model sample detection, notably improving detection of fake news articles generated by 20B parameter GPT-NeoX from 0.81 AUROC for the strongest zero-shot baseline to 0.95 AUROC for DetectGPT.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-mitchell23a, title = {{D}etect{GPT}: Zero-Shot Machine-Generated Text Detection using Probability Curvature}, author = {Mitchell, Eric and Lee, Yoonho and Khazatsky, Alexander and Manning, Christopher D and Finn, Chelsea}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {24950--24962}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/mitchell23a/mitchell23a.pdf}, url = {https://proceedings.mlr.press/v202/mitchell23a.html}, abstract = {The increasing fluency and widespread usage of large language models (LLMs) highlight the desirability of corresponding tools aiding detection of LLM-generated text. In this paper, we identify a property of the structure of an LLM’s probability function that is useful for such detection. Specifically, we demonstrate that text sampled from an LLM tends to occupy negative curvature regions of the model’s log probability function. Leveraging this observation, we then define a new curvature-based criterion for judging if a passage is generated from a given LLM. This approach, which we call DetectGPT, does not require training a separate classifier, collecting a dataset of real or generated passages, or explicitly watermarking generated text. It uses only log probabilities computed by the model of interest and random perturbations of the passage from another generic pre-trained language model (e.g., T5). We find DetectGPT is more discriminative than existing zero-shot methods for model sample detection, notably improving detection of fake news articles generated by 20B parameter GPT-NeoX from 0.81 AUROC for the strongest zero-shot baseline to 0.95 AUROC for DetectGPT.} }
Endnote
%0 Conference Paper %T DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature %A Eric Mitchell %A Yoonho Lee %A Alexander Khazatsky %A Christopher D Manning %A Chelsea Finn %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-mitchell23a %I PMLR %P 24950--24962 %U https://proceedings.mlr.press/v202/mitchell23a.html %V 202 %X The increasing fluency and widespread usage of large language models (LLMs) highlight the desirability of corresponding tools aiding detection of LLM-generated text. In this paper, we identify a property of the structure of an LLM’s probability function that is useful for such detection. Specifically, we demonstrate that text sampled from an LLM tends to occupy negative curvature regions of the model’s log probability function. Leveraging this observation, we then define a new curvature-based criterion for judging if a passage is generated from a given LLM. This approach, which we call DetectGPT, does not require training a separate classifier, collecting a dataset of real or generated passages, or explicitly watermarking generated text. It uses only log probabilities computed by the model of interest and random perturbations of the passage from another generic pre-trained language model (e.g., T5). We find DetectGPT is more discriminative than existing zero-shot methods for model sample detection, notably improving detection of fake news articles generated by 20B parameter GPT-NeoX from 0.81 AUROC for the strongest zero-shot baseline to 0.95 AUROC for DetectGPT.
APA
Mitchell, E., Lee, Y., Khazatsky, A., Manning, C.D. & Finn, C.. (2023). DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:24950-24962 Available from https://proceedings.mlr.press/v202/mitchell23a.html.

Related Material