Summary of Why Does Chatgpt “delve” So Much? Exploring the Sources Of Lexical Overrepresentation in Large Language Models, by Tom S. Juzek et al.

Why Does ChatGPT “Delve” So Much? Exploring the Sources of Lexical Overrepresentation in Large Language Models

by Tom S. Juzek, Zina B. Ward

First submitted to arxiv on: 16 Dec 2024

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 rapid changes in scientific English, particularly the increased use of words like “delve” and “intricate”. The authors develop a method to characterize these linguistic shifts, which reveals 21 focal words whose usage is likely driven by large language models (LLMs). The researchers explore why LLMs overuse these words, considering factors such as model architecture, algorithm choices, and training data. While they find some evidence that reinforcement learning from human feedback (RLHF) may contribute to lexical overrepresentation, their experimental results suggest participants react differently to certain focal words like “delve”. As LLMs become a driving force in global language change, understanding these potential sources of lexical overrepresentation is crucial.
Low	GrooveSquid.com (original content)	Low Difficulty Summary This paper looks at how scientific writing has changed recently. Scientists are using big computer models (LLMs) more often, and this might be why certain words, like “delve” and “intricate”, are being used a lot more. The researchers create a way to study these changes and find 21 important words that are probably caused by LLMs. They try to figure out why LLMs use these words so much, but they don’t find any clear answers. They do some experiments to see if people’s feedback helps explain the problem, but it’s not that simple. As computer models start to shape how we write and talk, understanding what’s going on is important.