Why Do Pretrained Language Models Help in Downstream Tasks? An Analysis of Head and Prompt Tuning
This provides incremental theoretical insights for NLP researchers into the mechanisms behind model adaptation, but it is limited to synthetic data and specific generative models.
The paper tackles the theoretical understanding of why pretrained language models aid downstream tasks by proposing an analysis framework linking pretraining and downstream tasks through latent variable generative models, showing that head tuning can solve tasks under non-degeneracy conditions and prompt tuning requires weaker conditions, with experiments on synthetic HMM data supporting these findings.
Pretrained language models have achieved state-of-the-art performance when adapted to a downstream NLP task. However, theoretical analysis of these models is scarce and challenging since the pretraining and downstream tasks can be very different. We propose an analysis framework that links the pretraining and downstream tasks with an underlying latent variable generative model of text -- the downstream classifier must recover a function of the posterior distribution over the latent variables. We analyze head tuning (learning a classifier on top of the frozen pretrained model) and prompt tuning in this setting. The generative model in our analysis is either a Hidden Markov Model (HMM) or an HMM augmented with a latent memory component, motivated by long-term dependencies in natural language. We show that 1) under certain non-degeneracy conditions on the HMM, simple classification heads can solve the downstream task, 2) prompt tuning obtains downstream guarantees with weaker non-degeneracy conditions, and 3) our recovery guarantees for the memory-augmented HMM are stronger than for the vanilla HMM because task-relevant information is easier to recover from the long-term memory. Experiments on synthetically generated data from HMMs back our theoretical findings.