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.
翻译:文字学习(ICL)使预先培训的语言模型能够在不经过任何梯度培训的情况下完成先前未见的任务,而无需经过任何梯度培训,作为投入的一部分,输入少量培训实例。ICL产生大量的计算、内存和存储费用,因为每次作出预测都涉及处理所有培训实例。PEFT(PEFT)(例如,适配模块、即时调整、稀疏更新方法等)提供了一个替代模式,其中培训了一套小型参数,以便能够执行新任务。在本文中,我们严格比较了少发的ICL和PEFT, 并表明后者提供了更好的准确性以及大大降低的计算费用。在前进的道路上,我们采用了一种新的PEEFT方法,称为(IA)$3美元,该方法通过学习的载体来激活尺度,同时只引入相对微小的数量的新参数。我们还提出了一个简单的方法,即T0模式称为T-Few,可以适用于新任务,而无需具体调整或修改的新任务。我们用SOFA-FS-FS-S-FS-FS-SUnalalalalalal ex Acal exalalal ex acal ex exal eximal exisal exit the exisal exisal exisal lapal exactal exital exital exit the ex ex exitalitalital exital 6-sal exital exital exital exitalitalitalitalitalutal 6-to ex exital exital exital exital-to the exital-to the exital exital exital exital-sal ex exital ex ex ex ex 6-to exital exital exital 6-toal exital-toal-to exital exital exital exital exital-toal-toal exital-toal-toal-toal exital-toal exital-toal ex ex