Evaluating Large Language Models on Controlled Generation Tasks
Computation and Language
2023-10-24 v1
Abstract
While recent studies have looked into the abilities of large language models in various benchmark tasks, including question generation, reading comprehension, multilingual and etc, there have been few studies looking into the controllability of large language models on generation tasks. We present an extensive analysis of various benchmarks including a sentence planning benchmark with different granularities. After comparing large language models against state-of-the-start finetuned smaller models, we present a spectrum showing large language models falling behind, are comparable, or exceed the ability of smaller models. We conclude that **large language models struggle at meeting fine-grained hard constraints**.
Cite
@article{arxiv.2310.14542,
title = {Evaluating Large Language Models on Controlled Generation Tasks},
author = {Jiao Sun and Yufei Tian and Wangchunshu Zhou and Nan Xu and Qian Hu and Rahul Gupta and John Frederick Wieting and Nanyun Peng and Xuezhe Ma},
journal= {arXiv preprint arXiv:2310.14542},
year = {2023}
}
Comments
EMNLP 2023