数据：人工智能的未来：法学硕士的飞速发展：到2030年，法学硕士可以在几个小时内完成长达一个月的任务

IF 2.4 4区工程技术 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

IEEE Spectrum Pub Date : 2025-07-08 DOI:10.1109/MSPEC.2025.11074452

Glenn Zorpette

引用次数: 0

摘要

对大型语言模型进行基准测试会带来一些不同寻常的挑战。首先，许多法学硕士的主要目的是提供与人类写作没有区别的引人注目的文本。而且，该任务的成功可能与传统上用于判断处理器性能的指标（如指令执行率）无关。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文本刊更多论文

The Data: The Future of AI: LLMs at Warp Speed: By 2030, LLMs May Tackle Monthlong Tasks in Hours

Benchmarking large language models presents some unusual challenges. For one, the main purpose of many LLMs is to provide compelling text that's indistinguishable from human writing. And success in that task may not correlate with metrics traditionally used to judge processor performance, such as instruction execution rate.

求助全文

通过发布文献求助，成功后即可免费获取论文全文。去求助

来源期刊

IEEE Spectrum 工程技术-工程：电子与电气

CiteScore

2.50

自引率

0.00%

发文量

254

审稿时长

4-8 weeks

期刊介绍： IEEE Spectrum Magazine, the flagship publication of the IEEE, explores the development, applications and implications of new technologies. It anticipates trends in engineering, science, and technology, and provides a forum for understanding, discussion and leadership in these areas. IEEE Spectrum is the world''s leading engineering and scientific magazine. Read by over 300,000 engineers worldwide, Spectrum provides international coverage of all technical issues and advances in computers, communications, and electronics. Written in clear, concise language for the non-specialist, Spectrum''s high editorial standards and worldwide resources ensure technical accuracy and state-of-the-art relevance.