Probing via Prompting

Abstract

Probing is a popular method to discern what linguistic information is contained in the representations of pre-trained language models. However, the mechanism of selecting the probe model has recently been subject to intense debate, as it is not clear if the probes are merely extracting information or modeling the linguistic property themselves. To address this challenge, this paper introduces a novel model-free approach to probing, by formulating probing as a prompting task. We conduct experiments on five probing tasks and show that our approach is comparable or better at extracting information than diagnostic probes while learning much less on its own. We further combine the probing via prompting approach with attention head pruning to analyze where the model stores the linguistic information in its architecture. We then examine the usefulness of a specific linguistic property for pre-training by removing the heads that are essential to that property and evaluating the resulting model's performance on language modeling. https://github.com/rycolab/ probing-via-prompting

Linguistic Profiling of a Neural Language Model

“…Understanding "how linguistic concepts that were common as features in NLP systems are captured in neural networks" (Belinkov and Glass, 2019) has been the focus of many studies in the recent NLP research. It has been extensively shown that pretrained Neural Language Models (NLMs) are able to capture syntax-and semantic-sensitive phenomena (Hewitt and Manning, 2019;Pimentel et al, 2020;Li et al, 2022) and that there is a correlation between the degree of linguistic knowledge and its ability to solve correctly a downstream task (Miaschi et al, 2020;Sarti et al, 2021), although it is still highly debated (Ravichander et al, 2021). However, it has also been demonstrated that introducing additional linguistic information (Wang et al, 2019b;Glavaš and Vulić, 2021) during the pre-training phase can enhance models' performances.…”

A Study on Accessing Linguistic Information in Pre-Trained Language Models by Using Prompts

“…The authors test a range of probes and downstream tasks at dozens of checkpoints. Li et al (2022) argue that prompting acts as a model-free probe, thus eliminating the distinction between what the model knows and what the probe learns. They compare prompting to linear regression and MLP probing.…”

Prompting PaLM for Translation: Assessing Strategies and Performance

“…Work in this area includes tools to facilitate development of handcrafted prompts (Strobelt et al, 2022;Bach et al, 2022); algorithms to find optimal prompts through gradient-guided search (Shin et al, 2020) or exhaustive search through labels (Schick and Schütze, 2021) or both labels and templates (Gao et al, 2021); as well as studies on the effect of example order (Kumar and Talukdar, 2021;Lu et al, 2022). Hard prompts have also been used to analyze model capabilities (Garg et al, 2022;Li et al, 2022a), the role of data (Singh et al, 2022), and the nature of prompting itself (Min et al, 2022).…”