Abstract
通用智能的”智能-目标”正交性及”工具性趋同”论点均要求通用智能的发展要智善结合。目前大语言模型在能力(智)方面发展迅速,但在更具挑战性的价值对齐(善)方面研究相对滞后。本综述将概述对齐的基本概念和必要性,简述其存在的社会和技术挑战,分析大语言模型对齐的主要技术路线和方法,探讨如何对大语言模型对齐进行评测,并对未来趋势进行展望。”- Anthology ID:
- 2023.ccl-2.7
- Volume:
- Proceedings of the 22nd Chinese National Conference on Computational Linguistics (Volume 2: Frontier Forum)
- Month:
- August
- Year:
- 2023
- Address:
- Harbin, China
- Editor:
- Jiajun Zhang
- Venue:
- CCL
- SIG:
- Publisher:
- Chinese Information Processing Society of China
- Note:
- Pages:
- 77–87
- Language:
- Chinese
- URL:
- https://aclanthology.org/2023.ccl-2.7
- DOI:
- Cite (ACL):
- Xiong Deyi. 2023. 大语言模型对齐:概念、挑战、路线、评测及趋势(Large Language Model Alignment: Concepts, Challenges, Roadmaps, Evaluations and Trends). In Proceedings of the 22nd Chinese National Conference on Computational Linguistics (Volume 2: Frontier Forum), pages 77–87, Harbin, China. Chinese Information Processing Society of China.
- Cite (Informal):
- 大语言模型对齐:概念、挑战、路线、评测及趋势(Large Language Model Alignment: Concepts, Challenges, Roadmaps, Evaluations and Trends) (Deyi, CCL 2023)
- PDF:
- https://preview.aclanthology.org/emnlp-22-attachments/2023.ccl-2.7.pdf