Reproduce “GPT-DETOX: An In-Context Learning-Based Paraphraser for Text Detoxification” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ali Pesaranghader, Nikhil Verma, Manasa Bharadwaj
GPT-DETOX, developed by LG Electronics Toronto AI Lab, presents an in-context learning framework leveraging GPT-3.5 Turbo for text detoxification. It effectively rewrites toxic text into non-toxic alternatives while preserving meaning, achieving state-of-the-art results on the ParaDetox dataset without model fine-tuning.
Key insights
- Few-shot prompting with relevant examples generally outperforms zero-shot prompting for text detoxification.
- Specialized example selection methods, particularly Context-Matching Example Selection (CMES) based on semantic similarity, enhance detoxification quality by providing more relevant in-context examples.
- An Ensemble In-Context Learning (EICL) approach, which combines outputs from multiple prompting strategies, substantially improves overall performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.