# LLM-based Rewriting of Inappropriate Argumentation using Reinforcement Learning from Machine Feedback

arXiv: https://arxiv.org/abs/2406.03363 · HTML: https://mlsanity.com/paper/2406.03363/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 39 of its 71 references that are in the map.
- First posted on arXiv 5 Jun 2024.

## Builds on (most cited first, 30 of 39)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [OPT: Open Pre-trained Transformer Language Models](https://mlsanity.com/paper/2205.01068/) (2022)
- [Self-Instruct: Aligning Language Models with Self-Generated Instructions](https://mlsanity.com/paper/2212.10560/) (2022)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Fine-Tuning Language Models from Human Preferences](https://mlsanity.com/paper/1909.08593/) (2019)
- [BLOOM: A 176B-Parameter Open-Access Multilingual Language Model](https://mlsanity.com/paper/2211.05100/) (2022)
- [Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classification](https://mlsanity.com/paper/1903.04561/) (2019)
- [Deep Reinforcement Learning for Dialogue Generation](https://mlsanity.com/paper/1606.01541/) (2016)
- [Dear Sir or Madam, May I introduce the GYAFC Dataset: Corpus, Benchmarks and Metrics for Formality Style Transfer](https://mlsanity.com/paper/1803.06535/) (2018)
- [Ex Machina: Personal Attacks Seen at Scale](https://mlsanity.com/paper/1610.08914/) (2016)
- [Automatically Neutralizing Subjective Bias in Text](https://mlsanity.com/paper/1911.09709/) (2019)
- [Politeness Transfer: A Tag and Generate Approach](https://mlsanity.com/paper/2004.14257/) (2020)
- [Better Rewards Yield Better Summaries: Learning to Summarise Without References](https://mlsanity.com/paper/1909.01214/) (2019)
- [A Study of Reinforcement Learning for Neural Machine Translation](https://mlsanity.com/paper/1808.08866/) (2018)
- [CoEdIT: Text Editing by Task-Specific Instruction Tuning](https://mlsanity.com/paper/2305.09857/) (2023)
- [A Dual Reinforcement Learning Framework for Unsupervised Text Style Transfer](https://mlsanity.com/paper/1905.10060/) (2019)
- [Unpaired Sentiment-to-Sentiment Translation: A Cycled Reinforcement Learning Approach](https://mlsanity.com/paper/1805.05181/) (2018)
- [Fighting Offensive Language on Social Media with Unsupervised Text Style Transfer](https://mlsanity.com/paper/1805.07685/) (2018)
- [Automatic Argument Quality Assessment -- New Datasets and Methods](https://mlsanity.com/paper/1909.01007/) (2019)
- [LEWIS: Levenshtein Editing for Unsupervised Text Style Transfer](https://mlsanity.com/paper/2105.08206/) (2021)
- [You Only Prompt Once: On the Capabilities of Prompt Learning on Large Language Models to Tackle Toxic Content](https://mlsanity.com/paper/2308.05596/) (2023)
- [Reinforcement Learning Based Text Style Transfer without Parallel Training Corpus](https://mlsanity.com/paper/1903.10671/) (2019)
- [Before Name-calling: Dynamics and Triggers of Ad Hominem Fallacies in Web Argumentation](https://mlsanity.com/paper/1802.06613/) (2018)
- [Civil Rephrases Of Toxic Texts With Self-Supervised Transformers](https://mlsanity.com/paper/2102.05456/) (2021)
- [Thank you BART! Rewarding Pre-Trained Models Improves Formality Style Transfer](https://mlsanity.com/paper/2105.06947/) (2021)

## Built on by (most cited first, 7 of 7)

- [Content Moderation by LLM: From Accuracy to Legitimacy](https://mlsanity.com/paper/2409.03219/) (2024)
- [Leanabell-Prover-V2: Verifier-integrated Reasoning for Formal Theorem Proving via Reinforcement Learning](https://mlsanity.com/paper/2507.08649/) (2025)
- [Mind the Style Gap: Meta-Evaluation of Style and Attribute Transfer Metrics](https://mlsanity.com/paper/2502.15022/) (2025)
- [Self-Adaptive Paraphrasing and Preference Learning for Improved Claim Verifiability](https://mlsanity.com/paper/2412.11653/) (2024)
- [Content Fuzzing for Escaping Information Cocoons on Digital Social Media](https://mlsanity.com/paper/2604.05461/) (2026)
- [Intention-Adaptive LLM Fine-Tuning for Text Revision Generation](https://mlsanity.com/paper/2602.00477/) (2026)
- [Teaching LLMs Human-Like Editing of Inappropriate Argumentation via Reinforcement Learning](https://mlsanity.com/paper/2604.12770/) (2026)

## Abstract (arXiv)

Ensuring that online discussions are civil and productive is a major challenge for social media platforms. Such platforms usually rely both on users and on automated detection tools to flag inappropriate arguments of other users, which moderators then review. However, this kind of post-hoc moderation is expensive and time-consuming, and moderators are often overwhelmed by the amount and severity of flagged content. Instead, a promising alternative is to prevent negative behavior during content creation. This paper studies how inappropriate language in arguments can be computationally mitigated. We propose a reinforcement learning-based rewriting approach that balances content preservation and appropriateness based on existing classifiers, prompting an instruction-finetuned large language model (LLM) as our initial policy. Unlike related style transfer tasks, rewriting inappropriate arguments allows deleting and adding content permanently. It is therefore tackled on document level rather than sentence level. We evaluate different weighting schemes for the reward function in both absolute and relative human assessment studies. Systematic experiments on non-parallel data provide evidence that our approach can mitigate the inappropriateness of arguments while largely preserving their content. It significantly outperforms competitive baselines, including few-shot learning, prompting, and humans.