Aligning Large Language Models by On-Policy Self-Judgment

Sangkyu Lee, Sungdong Kim, Ashkan Yousefpour, Minjoon Seo, Kang Min Yoo, Youngjae Yu


Abstract
Existing approaches for aligning large language models with human preferences face a trade-off that requires a separate reward model (RM) for on-policy learning. In this paper, we present a novel alignment framework, SELF-JUDGE that (1) does on-policy learning and 2) is parameter efficient, as it does not require an additional RM for evaluating the samples for on-policy learning. To this end, we propose Judge-augmented Supervised Fine-Tuning (JSFT) to train a single model to act as both a policy and a judge. Specifically, we view the pairwise judgment task, choosing the better response from a response pair, as a special case of the instruction-following task. The resulting model can judge preferences of on-the-fly responses from current policy initialized from itself. Experimental results show the efficacy of SELF-JUDGE, outperforming baselines in preference benchmarks. We also show that the rejecting sampling by itself can improve performance further without an additional evaluator.
Anthology ID:
2024.acl-long.617
Volume:
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
August
Year:
2024
Address:
Bangkok, Thailand
Editors:
Lun-Wei Ku, Andre Martins, Vivek Srikumar
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
11442–11459
Language:
URL:
https://aclanthology.org/2024.acl-long.617
DOI:
10.18653/v1/2024.acl-long.617
Bibkey:
Cite (ACL):
Sangkyu Lee, Sungdong Kim, Ashkan Yousefpour, Minjoon Seo, Kang Min Yoo, and Youngjae Yu. 2024. Aligning Large Language Models by On-Policy Self-Judgment. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 11442–11459, Bangkok, Thailand. Association for Computational Linguistics.
Cite (Informal):
Aligning Large Language Models by On-Policy Self-Judgment (Lee et al., ACL 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.acl-long.617.pdf