Summary of Hatecot: An Explanation-enhanced Dataset For Generalizable Offensive Speech Detection Via Large Language Models, by Huy Nghiem et al.
HateCOT: An Explanation-Enhanced Dataset for Generalizable Offensive Speech Detection via Large Language Models
by Huy Nghiem, Hal Daumé III
First submitted to arxiv on: 18 Mar 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Social and Information Networks (cs.SI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper introduces HateCOT, a large-scale English dataset for detecting offensive content on social media. The dataset features over 52,000 samples from diverse sources, each with an explanation generated by GPT-3.5Turbo and curated by humans. The authors demonstrate that pretraining on HateCOT improves the performance of open-source Large Language Models (LLMs) in both zero-shot and few-shot settings for detecting offensive content, despite differences in domain and task. Additionally, HateCOT enables effective K-shot fine-tuning of LLMs with limited data and improves the quality of their explanations, as confirmed by human evaluation. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make social media safer by creating a big dataset to detect bad content. The dataset has lots of examples from different places online, each with an explanation generated by a special AI tool and checked by humans. The research shows that using this dataset can help special language models do better at finding bad content, even when they’re not trained specifically for that task. It also helps make those language models give better explanations. |
Keywords
» Artificial intelligence » Few shot » Fine tuning » Gpt » Pretraining » Zero shot
