Knowledge Commons of Institute of Automation,CAS
Token-level Direct Preference Optimization | |
Zeng,Yongcheng1; Liu,Guoqing2![]() ![]() | |
2024 | |
会议名称 | 2024 42nd International Conference on Machine Learning(ICML) |
会议日期 | 2024/7/21-27 |
会议地点 | Vienna, Austria |
摘要 | Fine-tuning pre-trained Large Language Models (LLMs) is essential to align them with human values and intentions. This process often uti- lizes methods like pairwise comparisons and KL divergence against a reference LLM, focusing on the evaluation of full answers generated by the models. However, the generation of these responses occurs in a token level, following a sequential, auto-regressive fashion. In this pa- per, we introduce Token-level Direct Preference Optimization (TDPO), a novel approach to align LLMs with human preferences by optimizing pol- icy at the token level. Unlike previous methods, which face challenges in divergence efficiency, TDPO incorporates forward KL divergence con- straints for each token, improving alignment and diversity. Utilizing the Bradley-Terry model for a token-based reward system, TDPO enhances the regulation of KL divergence, while preserv- ing simplicity without the need for explicit re- ward modeling. Experimental results across vari- ous text tasks demonstrate TDPO’s superior per- formance in balancing alignment with genera- tion diversity. Notably, fine-tuning with TDPO strikes a better balance than DPO in the controlled sentiment generation and single-turn dialogue datasets, and significantly improves the quality of generated responses compared to both DPO and PPO-based RLHF methods. Our code is open- sourced at https://github.com/Vance0124/Token- level-Direct-Preference-Optimization. |
学科门类 | 工学 ; 工学::计算机科学与技术(可授工学、理学学位) |
收录类别 | EI |
语种 | 英语 |
是否为代表性论文 | 是 |
七大方向——子方向分类 | 决策智能理论与方法 |
国重实验室规划方向分类 | 其他 |
是否有论文关联数据集需要存交 | 否 |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/57249 |
专题 | 复杂系统认知与决策实验室_群体决策智能团队 |
通讯作者 | Zhang,Haifeng; Wang,Jun |
作者单位 | 1.Institute of Automation, Chinese Academy of Sciences 2.Microsoft Research AI4Science 3.University College London |
第一作者单位 | 中国科学院自动化研究所 |
通讯作者单位 | 中国科学院自动化研究所 |
推荐引用方式 GB/T 7714 | Zeng,Yongcheng,Liu,Guoqing,Ma,Weiyu,et al. Token-level Direct Preference Optimization[C],2024. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
Token-level Direct P(883KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论