Unveiling the Lexical Sensitivity of LLMs: Combinatorial Optimization for Prompt Enhancement

Pengwei Zhan, Zhen Xu, Qian Tan, Jie Song, Ru Xie


Abstract
Large language models (LLMs) demonstrate exceptional instruct-following ability to complete various downstream tasks. Although this impressive ability makes LLMs flexible task solvers, their performance in solving tasks also heavily relies on instructions. In this paper, we reveal that LLMs are over-sensitive to lexical variations in task instructions, even when the variations are imperceptible to humans. By providing models with neighborhood instructions, which are closely situated in the latent representation space and differ by only one semantically similar word, the performance on downstream tasks can be vastly different. Following this property, we propose a black-box Combinatorial Optimization framework for Prompt Lexical Enhancement (COPLE). COPLE performs iterative lexical optimization according to the feedback from a batch of proxy tasks, using a search strategy related to word influence. Experiments show that even widely-used human-crafted prompts for current benchmarks suffer from the lexical sensitivity of models, and COPLE recovers the declined model ability in both instruct-following and solving downstream tasks.
Anthology ID:
2024.emnlp-main.295
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:
5128–5154
Language:
URL:
https://aclanthology.org/2024.emnlp-main.295/
DOI:
10.18653/v1/2024.emnlp-main.295
Bibkey:
Cite (ACL):
Pengwei Zhan, Zhen Xu, Qian Tan, Jie Song, and Ru Xie. 2024. Unveiling the Lexical Sensitivity of LLMs: Combinatorial Optimization for Prompt Enhancement. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 5128–5154, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
Unveiling the Lexical Sensitivity of LLMs: Combinatorial Optimization for Prompt Enhancement (Zhan et al., EMNLP 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.emnlp-main.295.pdf