Beyond Preserved Accuracy: Evaluating Loyalty and Robustness of BERT Compression

By Canwen Xu, Wangchunshu Zhou, Tao Ge, Kelvin J. Xu, Julian McAuley, Furu Wei. 2021

47 citations recorded by the source corpus.

Recent studies on compression of pretrained language models (e.g., BERT) usually use preserved accuracy as the metric for evaluation. In this paper, we propose two new metrics, label loyalty and probability loyalty that measure how closely a compressed model (i.e., student) mimics the original model (i.e., teacher). We also explore the effect of compression with regard to robustness under adversarial attacks. We benchmark quantization, pruning, knowledge distillation and progressive module replacing with loyalty and robustness. By combining multiple compression techniques, we provide a practical strategy to achieve better accuracy, loyalty and robustness.

Subfields: Artificial Intelligence, Efficiency & Compression

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Efficiency & Compression

Citations in the corpus

“I’m sorry to hear that”: Finding New Biases in Language Models with a Holistic Descriptor Dataset (2022, methodology): “In particular, many works focus on generative models (Dinan et al., 2020a,b; Xu et al., 2021b; Kirk et al., 2021; Sheng et al., 2021b; Nozza et al., 2021; Renduchintala et al., 2021; Baheti et al., 2021; Perez et al., 2022), which are well known to pose unique challenges for automatic evaluation…”
A Survey on Dynamic Neural Networks for Natural Language Processing (2022, background): “…static model acceleration, which is often referred to as model compression , including knowledge distillation, weight sharing, pruning and quantization (Sanh et al., 2019; Xu et al., 2020; Lan et al., 2020; Zafrir et al., 2019; Xu et al., 2021) (etc., see another survey (Xu and McAuley, 2022)).”
"I'm sorry to hear that": finding bias in language models with a holistic descriptor dataset (2022, background): “…empathy, and knowledge (Zhang et al., 2018; Rashkin et al., 2019; Dinan et al., 2018; Smith et al., 2020b; Roller et al., 2021), recall of past conversation details across multiple sessions (Xu et al., 2021a), and the ability to retrieve factual information from the internet (Komeili et al., 2021).”
Intriguing Properties of Compression on Multilingual Models (2022, background): “Interestingly, our ﬁndings are consistent across all perturbation types as their scope increases from languages ( in-language ) to scripts ( in-script ) and families ( in-family ).”