Scholarlink

This question describes Large Language Models (such as ChatGPT). It includes a discussion of how they work and the kind of information they use (and that they ignore) in writing texts. It further discusses the kinds of errors is likely to make and how they affect the evaluation and use of LLM-written documents. In particular, we argue that LLMs do not answer questions, they just produce documents that look like answers to the question.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

How Large Language Models Work

  • 章节

  • 发表日期：

    2025年

查看全文
  • Patrick Juola

摘要

This question describes Large Language Models (such as ChatGPT). It includes a discussion of how they work and the kind of information they use (and that they ignore) in writing texts. It further discusses the kinds of errors is likely to make and how they affect the evaluation and use of LLM-written documents. In particular, we argue that LLMs do not answer questions, they just produce documents that look like answers to the question.