FIRST: Teach A Reliable Large Language Model Through Efficient Trustworthy Distillation

KaShun Shum, Minrui Xu, Jianshu Zhang, Zixin Chen, Shizhe Diao, Hanze Dong, Jipeng Zhang, Muhammad Raza


Abstract
Large language models (LLMs) have become increasingly prevalent in our daily lives, leading to an expectation for LLMs to be trustworthy —- both accurate and well-calibrated (the prediction confidence should align with its ground truth correctness likelihood). Nowadays, fine-tuning has become the most popular method for adapting a model to practical usage by significantly increasing accuracy on downstream tasks. Despite the great accuracy it achieves, we found fine-tuning is still far away from satisfactory trustworthiness due to “tuning-induced mis-calibration”. In this paper, we delve deeply into why and how mis-calibration exists in fine-tuned models, and how distillation can alleviate the issue. Then we further propose a brand new method named Efficient Trustworthy Distillation (FIRST), which utilizes a small portion of teacher’s knowledge to obtain a reliable language model in a cost-efficient way. Specifically, we identify the “concentrated knowledge” phenomenon during distillation, which can significantly reduce the computational burden. Then we apply a “trustworthy maximization” process to optimize the utilization of this small portion of concentrated knowledge before transferring it to the student. Experimental results demonstrate the effectiveness of our method, where better accuracy (+2.3%) and less mis-calibration (-10%) are achieved on average across both in-domain and out-of-domain scenarios, indicating better trustworthiness.
Anthology ID:
2024.emnlp-main.703
Volume:
Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
12646–12659
Language:
URL:
https://aclanthology.org/2024.emnlp-main.703
DOI:
Bibkey:
Cite (ACL):
KaShun Shum, Minrui Xu, Jianshu Zhang, Zixin Chen, Shizhe Diao, Hanze Dong, Jipeng Zhang, and Muhammad Raza. 2024. FIRST: Teach A Reliable Large Language Model Through Efficient Trustworthy Distillation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 12646–12659, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
FIRST: Teach A Reliable Large Language Model Through Efficient Trustworthy Distillation (Shum et al., EMNLP 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.emnlp-main.703.pdf