Challenges and Applications of Large Language Models
Jean Kaddour University College London Joshua Harris UK Health Security Agency Maximilian Mozes University College London Herbie Bradley EleutherAI University of Cambridge Stability AI Roberta Raileanu Meta AI Research Robert McHardy InstaDeep
Abstract
Large Language Models (LLMs) went from non-existent to ubiquitous in the machine learning discourse within a few years. Due to the fast pace of the field, it is difficult to identify the remaining challenges and already fruitful application areas. In this paper, we aim to establish a systematic set of open problems and application successes so that ML researchers can comprehend the field’s current state more quickly and become productive. 11footnotetext: Equal contribution. 22footnotetext: {jean.kaddour,robert.mchardy}.20@ucl.ac.uk, joshua.harris@ukhsa.gov.uk
中文速览
大型语言模型（LLM）近年来在机器学习领域迅速普及，但该领域发展太快，研究者很难快速摸清哪些问题已经解决、哪些挑战仍悬而未决。这篇综述系统梳理了LLM面临的核心挑战——涵盖数据集质量、分词依赖、高昂训练成本、模型幻觉与偏见、推理能力不足等——并同时盘点了LLM在聊天机器人、计算生物学、编程辅助、法律、医学、机器人等十余个应用领域的落地进展。作者将挑战按"设计""行为""科学"三大类归纳，并分析了各类挑战如何制约实际应用。这份全景式梳理的价值在于，能帮助技术研究者迅速定位当前LLM的能力边界与研究空白，少走弯路，更快投入到有价值的方向上。
原文 arXiv:2307.10169;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2307.10169v1