Summary of The Gus Framework: Benchmarking Social Bias Classification with Discriminative (encoder-only) and Generative (decoder-only) Language Models, by Maximus Powers et al.
The GUS Framework: Benchmarking Social Bias Classification with Discriminative (Encoder-Only) and Generative (Decoder-Only) Language Models
by Maximus Powers, Shaina Raza, Alex Chang, Umang Mavani, Harshitha Reddy Jonala, Ansh Tiwari, Hua Wei
First submitted to arxiv on: 10 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
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 called Generalizations, Unfairness, and Stereotypes (GUS) to detect social bias in text. The GUS framework focuses on three key linguistic components underlying social bias: generalizations, unfairness, and stereotypes. To create a comprehensive synthetic dataset, the authors employ a semi-automated approach and verify it with humans to maintain ethical standards. This dataset enables robust multi-label token classification. The authors combine discriminative (encoder-only) models and generative (auto-regressive large language models) to identify biased entities in text. Through extensive experiments, they demonstrate that encoder-only models are effective for this complex task, often outperforming state-of-the-art methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us understand how to detect social bias in writing. Right now, we only have simple ways to spot biased content, which can be wrong and hurtful. The authors created a new framework called GUS that looks at three things: generalizations, unfairness, and stereotypes. They made a special dataset that helps computers learn to recognize these biases. The authors tested different computer models and found that some are better than others at finding biased words or phrases. This research can help us make fairer decisions in many areas of life. |
Keywords
» Artificial intelligence » Classification » Encoder » Token
