EMNLP Findings2022
SMASH: Improving SMAll Language Models' Few-SHot Ability with Prompt-Based Distillation
Abstract
Large-scale language models coupled with prompts have shown remarkable performance on few-shot learning. However, through systematic experiments, we find that the few-shot performance of small language models is poor, and using prompts on them brings fewer improvements than on larger ones. In this paper, we propose SMASH, an approach to improve SMAll language models' few-SHot ability by training on intermediate tasks before promptbased fine-tuning on downstream tasks. We design intermediate tasks for sentence-pair tasks and sentiment classification tasks by creating training examples with prompt templates similar to downstream tasks using sentences sampled from a large-scale unsupervised corpus, and apply knowledge distillation to distill from outputs of larger pre-trained models as the training objective. We conduct extensive experiments and show that SMASH can make a 6layer DistilRoBRETa-base achieve comparable performance on few-shot datasets with a 12layer RoBERTa-base at a low cost. 1
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
1 top-tier paper has cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Differentiable Prompt Makes Pre-trained Language Models Better Few-shot Learners cites AutoPrompt: Eliciting Knowledge from Language Models with Automatically Generated Prompts
- Differentiable Prompt Makes Pre-trained Language Models Better Few-shot Learners cites Language Models are Few-Shot Learners
- AutoPrompt: Eliciting Knowledge from Language Models with Automatically Generated Prompts cites Language Models are Few-Shot Learners
Cited by top-tier papers1
Builds on12
- Language Models are Few-Shot LearnersTom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah et al.NeurIPS 2020 · 64,255 citations
- AutoPrompt: Eliciting Knowledge from Language Models with Automatically Generated PromptsTaylor Shin, Yasaman Razeghi, Robert L. Logan IV, Eric Wallace et al.EMNLP 2020 · 1,162 citations
- Climbing towards NLU: On Meaning, Form, and Understanding in the Age of DataEmily M. Bender, Alexander KollerACL 2020 · 914 citations
- MiniLMv2: Multi-Head Self-Attention Relation Distillation for Compressing Pretrained TransformersWenhui Wang, Hangbo Bao, Shaohan Huang, Li Dong et al.ACL Findings 2021 · 260 citations
- Differentiable Prompt Makes Pre-trained Language Models Better Few-shot LearnersNingyu Zhang, Luoqiu Li, Xiang Chen, Shumin Deng et al.ICLR 2022 · 205 citations
Related papers
- Adversarial Robustness of Prompt-based Few-Shot Learning for Natural Language UnderstandingVenkata Prabhakara Sarath Nookala, Gaurav Verma, Subhabrata Mukherjee, Srijan KumarACL Findings 2023 · 3 citations
- Making Pre-trained Language Models Better Few-shot LearnersTianyu Gao, Adam Fisch, Danqi ChenACL 2021
- Towards Unified Prompt Tuning for Few-shot Text ClassificationJianing Wang, Chengyu Wang, Fuli Luo, Chuanqi Tan et al.EMNLP Findings 2022 · 37 citations
- Revisiting Self-training for Few-shot Learning of Language ModelYiming Chen, Yan Zhang, Chen Zhang, Grandee Lee et al.EMNLP 2021 · 35 citations
- Reordering Examples Helps during Priming-based Few-Shot LearningSawan Kumar, Partha P. TalukdarACL Findings 2021 · 49 citations
