Summary of Shortcut Learning in In-context Learning: a Survey, by Rui Song et al.
Shortcut Learning in In-Context Learning: A Survey
by Rui Song, Yingji Li, Lida Shi, Fausto Giunchiglia, Hao Xu
First submitted to arxiv on: 4 Nov 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 The abstract discusses the phenomenon of “shortcut learning” in large language models (LLMs), where they employ simple decision rules instead of robust ones, hindering their generalization and robustness. The paper provides a novel perspective on shortcut learning in In-Context Learning (ICL) tasks, exploring types of shortcuts, causes, benchmarks, and strategies for mitigating them. It also summarizes unresolved issues and outlines the future research landscape. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models are getting better at understanding human language, but they still have some major limitations. One problem is that they often use “shortcuts” to make decisions instead of really understanding what’s going on. This makes it hard for them to work well in real-world situations. The authors of this paper want to understand more about these shortcuts and how we can fix the problems they cause. |
Keywords
» Artificial intelligence » Generalization
