Summary of Robustness Of Llms to Perturbations in Text, by Ayush Singh et al.
Robustness of LLMs to Perturbations in Text
by Ayush Singh, Navpreet Singh, Shubham Vatsal
First submitted to arxiv on: 12 Jul 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 investigates the resilience of Large Language Models (LLMs) against noise in real-world text data. The authors artificially introduce varying levels of noise into a diverse set of datasets and evaluate LLMs’ robustness against corrupt text variations. Contrary to popular beliefs, generative LLMs are found to be surprisingly robust to noisy perturbations in text. This is a departure from pre-trained models like BERT or RoBERTa, whose performance has been shown to be sensitive to deteriorating noisy text. The authors also test LLMs’ resilience on multiple real-world benchmarks that mimic commonly found errors in the wild. With minimal prompting, LLMs achieve a new state-of-the-art on benchmark tasks of Grammar Error Correction (GEC) and Lexical Semantic Change (LSC). To empower future research, the authors release a dataset annotated by humans stating their preference for LLM vs. human-corrected outputs along with the code to reproduce results. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how well large language models can handle mistakes in real-world text data. Most language processing systems assume that the text is perfect, but this isn’t always true. The authors test these models by adding different levels of errors into a variety of datasets and see how they perform. They find that some types of models are very good at handling mistakes and even do better than expert-corrected versions. This is important because it shows that language models can be used in real-world applications where text may not be perfect. |
Keywords
» Artificial intelligence » Bert » Prompting
