Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learning

TL;DR

This paper demonstrates that parameter-efficient fine-tuning (PEFT) methods outperform in-context learning (ICL) in accuracy and cost, introduces a new PEFT method (IA)$^3$, and presents T-Few, a versatile approach achieving super-human performance on unseen tasks.

Contribution

The paper provides a comprehensive comparison of PEFT and ICL, introduces the IA)$^3$ method, and proposes T-Few for zero-shot task generalization with state-of-the-art results.

Findings

PEFT outperforms ICL in accuracy and efficiency.

The new IA)$^3$ method improves performance with minimal additional parameters.

T-Few achieves super-human performance on unseen tasks, surpassing state-of-the-art.

Abstract

Few-shot in-context learning (ICL) enables pre-trained language models to perform a previously-unseen task without any gradient-based training by feeding a small number of training examples as part of the input. ICL incurs substantial computational, memory, and storage costs because it involves processing all of the training examples every time a prediction is made. Parameter-efficient fine-tuning (PEFT) (e.g. adapter modules, prompt tuning, sparse update methods, etc.) offers an alternative paradigm where a small set of parameters are trained to enable a model to perform the new task. In this paper, we rigorously compare few-shot ICL and PEFT and demonstrate that the latter offers better accuracy as well as dramatically lower computational costs. Along the way, we introduce a new PEFT method called (IA)$^3$ that scales activations by learned vectors, attaining stronger performance while only introducing a relatively tiny amount of new parameters. We also propose a simple recipe based on the T0 model called T-Few that can be applied to new tasks without task-specific tuning or modifications. We validate the effectiveness of T-Few on completely unseen tasks by applying it to the RAFT benchmark, attaining super-human performance for the first time and outperforming the state-of-the-art by 6% absolute. All of the code used in our experiments is publicly available.