GLTR: Statistical Detection and Visualization of Generated Text

By Sebastian Gehrmann, Hendrik Strobelt, Alexander M. Rush. 2019

852 citations recorded by the source corpus.

The rapid improvement of language models has raised the specter of abuse of text generation systems. This progress motivates the development of simple methods for detecting generated text that can be used by non-experts. In this work, we introduce GLTR, a tool to support humans in detecting whether a text was generated by a model. GLTR applies a suite of baseline statistical methods that can detect generation artifacts across multiple sampling schemes. In a human-subjects study, we show that the annotation scheme provided by GLTR improves the human detection-rate of fake text from 54% to 72% without any prior training. GLTR is open-source and publicly deployed, and has already been widely used to detect generated outputs.

Subfields: Artificial Intelligence, Sociology and Political Science, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Sociology and Political Science

In Safety, Robustness & Interpretability

Citations in the corpus

Automatic Detection of Generated Text is Easiest when Humans are Fooled (2019, methodology): “Kreps et al. (2020), Gehrmann et al. (2019), and others have stressed the importance of humans being able to identify fake content on the web.”
Release Strategies and the Social Impacts of Language Models (2019, methodology): “on GLTR [26], which shows these probabilities to humans in a friendly interface.”
Repairing the Cracked Foundation: A Survey of Obstacles in Evaluation Practices for Generated Text (2022, methodology): “While some work exists that aims to standardize extrinsic evaluations (e.g., Mani et al., 1999; Gehrmann et al., 2019a), the design space is much larger.”
Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Survey (2022, background): “And finally, (4) analysis and visualization tools that allow developers to debug how LMs make predictions (Gehrmann et al., 2019).”
Self-critiquing models for assisting human evaluators (2022, background): “[GSR19] use language models to help humans discriminate whether text was generated by a model.”
GLM-130B: An Open Bilingual Pre-trained Model (2022, background): “Conversely, releasing an open LLM can provide access and transparency to all the researchers and promote the research to reduce the potential harm of LLMs, like algorithms to identify the synthetic text Gehrmann et al. (2019) or detect fake news Li et al. (2021).”
The Gradient of Generative AI Release: Methods and Considerations (2023, methodology): “Approaches to detection can be tailored to modality [42] such as text [35], and include human annotation [29].”
Dats Wassup!!: Investigating African-American Vernacular English in Transformer-Based Text Generation (2020, background): “Coupled with concerns that NLG tools can be used for generating fake news (Gehrmann et al., 2019) or impersonating internet”
Real or Fake? Learning to Discriminate Machine from Human Generated Text (2019, background): “Similarly, Gehrmann et al. (2019) (GLTR) assume knowledge of the generator; these Authors say “ We further hypothesize that these methods generalize to black-box scenarios, as long as the fake text follows a similar sampling assumption and is generated by a large language model ”; our work answers…”
Automatic Detection of Machine Generated Text: A Critical Survey (2020, methodology): “The movie review (12) switches the focus of the discourse between actors (Sophia and Duchovny) and story line in an incoherent fashion, which violates the theory of centering in discourse analysis (Grosz et al., 1995; Gehrmann et al., 2019).”
Improved Natural Language Generation via Loss Truncation (2020, background): “Distin-guishability has recently gained attention as a way to learn and evaluate models based on both sample quality and diversity (Hashimoto et al., 2019; Zhou et al., 2019; Zellers et al., 2019; Gehrmann et al., 2019).”
Limits of Detecting Text Generated by Large-Scale Language Models (2020, background): “The alarming consequences of such machine-generated misinformation present an urgent need to discern fake content from genuine, as it is becoming more and more difficult for people to do so without cognitive support tools [14].”
DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature (2023, methodology): “The most common method for zero-shot machine-generated text detection is evaluating the average per-token log probability of the generated text and thresholding (Solaiman et al., 2019; Gehrmann et al., 2019; Ippolito et al., 2020).”
Can AI-Generated Text be Reliably Detected? (2023, methodology): “Non-watermarking detectors such as trained classiﬁers [3], retrieval-based detectors [4], and zero-shot classiﬁers [2, 24, 23, 22] use the presence of LLM-speciﬁc signatures in AI-generated texts for their detection.”