Summary of Eagle: a Domain Generalization Framework For Ai-generated Text Detection, by Amrita Bhattacharjee et al.
EAGLE: A Domain Generalization Framework for AI-generated Text Detection
by Amrita Bhattacharjee, Raha Moraffah, Joshua Garland, Huan Liu
First submitted to arxiv on: 23 Mar 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
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 This paper proposes a novel framework, EAGLE, for detecting AI-generated text from unseen target generators using large language models (LLMs). Existing supervised detectors perform well on older LLMs but require new labeled training data for newer models. To address this issue, EAGLE leverages available labeled data from older LLMs and learns domain-invariant features through self-supervised contrastive learning and domain adversarial training. The proposed framework demonstrates impressive performance in detecting text generated by unseen target generators, including recent state-of-the-art models like GPT-4 and Claude, achieving detection scores within 4.7% of a fully supervised detector. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research focuses on making it easier to identify text written by artificial intelligence (AI) language models. Right now, there are many different AI models that can generate text, but most detectors work well only for older models. To solve this problem, the researchers created a new method called EAGLE that uses existing data from older models and learns ways to recognize features that are consistent across all AI-generated text. This allows EAGLE to accurately identify text generated by new, unseen AI models with high accuracy. |
Keywords
* Artificial intelligence * Claude * Gpt * Self supervised * Supervised
