‘AI is, at its heart and best, a tool’: Students, Faculty Weigh in on the Use of Large Language Models to Read Academic Papers
Publish Date: 2026-05-10 22:50:00
Source Domain: www.cornellsun.com
Here’s a summary of the article, formatted as an unordered list with 6 key points:
- The Study: A study conducted by Cornell physicists and Google researchers evaluated the performance of six Large Language Models (LLMs) including ChatGPT, Claude, and Google Gemini in their ability to read scientific literature at a specialist level.
- Findings: Some LLMs performed better than others, revealing gaps in the current models’ abilities. NotebookLM and the RAG system performed best, pulling answers from more reputable sources.
- Areas for Improvement: The study highlighted areas where future AI model improvements are needed, particularly in understanding and utilizing high-quality references and data visualization.
- Student Use: Some students are using LLMs to aid in reading scientific papers, citing the tools as helpful for gaining a cursory understanding of a subject. However, caution is advised due to potential reliability issues.
- Limitations: LLMs are prone to pulling information from non-peer-reviewed sources which can be unreliable. Despite AI’s assistance, students still need to critically evaluate the scientific literature they use.
- Expert Recommendation: Prof. Eunah Ah-Kim emphasizes the importance of critically reading scientific literature and understanding the evidence, even while acknowledging the valuable role AI can play in making research endeavors more collaborative.
