# Unsupervised Text Style Transfer using Language Models as Discriminators

arXiv: https://arxiv.org/abs/1805.11749 · HTML: https://mlsanity.com/paper/1805.11749/

## Key facts

- Cited by 70 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 30 May 2018.

## Built on by (most cited first, 30 of 70)

- [RADAR: Robust AI-Text Detection via Adversarial Learning](https://mlsanity.com/paper/2307.03838/) (2023)
- [Mind the Style of Text! Adversarial and Backdoor Attacks Based on Text Style Transfer](https://mlsanity.com/paper/2110.07139/) (2021)
- [In-Context Impersonation Reveals Large Language Models' Strengths and Biases](https://mlsanity.com/paper/2305.14930/) (2023)
- [Break-It-Fix-It: Unsupervised Learning for Program Repair](https://mlsanity.com/paper/2106.06600/) (2021)
- [BadPrompt: Backdoor Attacks on Continuous Prompts](https://mlsanity.com/paper/2211.14719/) (2022)
- [A Dense Reward View on Aligning Text-to-Image Diffusion with Preference](https://mlsanity.com/paper/2402.08265/) (2024)
- [A Causal Lens for Controllable Text Generation](https://mlsanity.com/paper/2201.09119/) (2022)
- [MuseMorphose: Full-Song and Fine-Grained Piano Music Style Transfer with One Transformer VAE](https://mlsanity.com/paper/2105.04090/) (2021)
- [The Threat of Offensive AI to Organizations](https://mlsanity.com/paper/2106.15764/) (2021)
- [Gradient-Based Language Model Red Teaming](https://mlsanity.com/paper/2401.16656/) (2024)
- [Civil Rephrases Of Toxic Texts With Self-Supervised Transformers](https://mlsanity.com/paper/2102.05456/) (2021)
- [Preference-grounded Token-level Guidance for Language Model Fine-tuning](https://mlsanity.com/paper/2306.00398/) (2023)
- [Text Style Transfer for Bias Mitigation using Masked Language Modeling](https://mlsanity.com/paper/2201.08643/) (2022)
- [Controllable Data Generation by Deep Learning: A Review](https://mlsanity.com/paper/2207.09542/) (2022)
- [Machine Translation into Low-resource Language Varieties](https://mlsanity.com/paper/2106.06797/) (2021)
- [Improving Fake News Detection of Influential Domain via Domain- and Instance-Level Transfer](https://mlsanity.com/paper/2209.08902/) (2022)
- [A Paradigm Shift: The Future of Machine Translation Lies with Large Language Models](https://mlsanity.com/paper/2305.01181/) (2023)
- [Deep Extrapolation for Attribute-Enhanced Generation](https://mlsanity.com/paper/2107.02968/) (2021)
- [A Call for Standardization and Validation of Text Style Transfer Evaluation](https://mlsanity.com/paper/2306.00539/) (2023)
- [Inference Time Style Control for Summarization](https://mlsanity.com/paper/2104.01724/) (2021)
- [Counterfactual Generation with Identifiability Guarantees](https://mlsanity.com/paper/2402.15309/) (2024)
- [Countering Online Hate Speech: An NLP Perspective](https://mlsanity.com/paper/2109.02941/) (2021)
- [Style Pooling: Automatic Text Style Obfuscation for Improved Classification Fairness](https://mlsanity.com/paper/2109.04624/) (2021)
- [Summarize and Generate to Back-translate: Unsupervised Translation of Programming Languages](https://mlsanity.com/paper/2205.11116/) (2022)
- [Segmenting Text and Learning Their Rewards for Improved RLHF in Language Model](https://mlsanity.com/paper/2501.02790/) (2025)
- [Duplex Sequence-to-Sequence Learning for Reversible Machine Translation](https://mlsanity.com/paper/2105.03458/) (2021)
- [Toward a `Standard Model' of Machine Learning](https://mlsanity.com/paper/2108.07783/) (2021)
- [NAST: A Non-Autoregressive Generator with Word Alignment for Unsupervised Text Style Transfer](https://mlsanity.com/paper/2106.02210/) (2021)
- [Efficient Reinforcement Learning for Unsupervised Controlled Text Generation](https://mlsanity.com/paper/2204.07696/) (2022)
- [Towards Robust and Semantically Organised Latent Representations for Unsupervised Text Style Transfer](https://mlsanity.com/paper/2205.02309/) (2022)

## Abstract (arXiv)

Binary classifiers are often employed as discriminators in GAN-based unsupervised style transfer systems to ensure that transferred sentences are similar to sentences in the target domain. One difficulty with this approach is that the error signal provided by the discriminator can be unstable and is sometimes insufficient to train the generator to produce fluent language. In this paper, we propose a new technique that uses a target domain language model as the discriminator, providing richer and more stable token-level feedback during the learning process. We train the generator to minimize the negative log likelihood (NLL) of generated sentences, evaluated by the language model. By using a continuous approximation of discrete sampling under the generator, our model can be trained using back-propagation in an end- to-end fashion. Moreover, our empirical results show that when using a language model as a structured discriminator, it is possible to forgo adversarial steps during training, making the process more stable. We compare our model with previous work using convolutional neural networks (CNNs) as discriminators and show that our approach leads to improved performance on three tasks: word substitution decipherment, sentiment modification, and related language translation.