@inproceedings{zhao-etal-2022-measuring,
title = "On Measuring the Intrinsic Few-Shot Hardness of Datasets",
author = "Zhao, Xinran and
Murty, Shikhar and
Manning, Christopher",
editor = "Goldberg, Yoav and
Kozareva, Zornitsa and
Zhang, Yue",
booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing",
month = dec,
year = "2022",
address = "Abu Dhabi, United Arab Emirates",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2022.emnlp-main.262/",
doi = "10.18653/v1/2022.emnlp-main.262",
pages = "3955--3963",
abstract = "While advances in pre-training have led to dramatic improvements in few-shot learning of NLP tasks, there is limited understanding of what drives successful few-shot adaptation in datasets. In particular, given a new dataset and a pre-trained model, what properties of the dataset make it few-shot learnable, and are these properties independent of the specific adaptation techniques used? We consider an extensive set of recent few-shot learning methods and show that their performance across a large number of datasets is highly correlated, showing that few-shot hardness may be intrinsic to datasets, for a given pre-trained model. To estimate intrinsic few-shot hardness, we then propose a simple and lightweight metric called Spread that captures the intuition that few-shot learning is made possible by exploiting feature-space invariances between training and test samples. Our metric better accounts for few-shot hardness compared to existing notions of hardness and is {\textasciitilde}8-100x faster to compute."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhao-etal-2022-measuring">
<titleInfo>
<title>On Measuring the Intrinsic Few-Shot Hardness of Datasets</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xinran</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shikhar</namePart>
<namePart type="family">Murty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christopher</namePart>
<namePart type="family">Manning</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2022-12</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yoav</namePart>
<namePart type="family">Goldberg</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zornitsa</namePart>
<namePart type="family">Kozareva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Abu Dhabi, United Arab Emirates</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>While advances in pre-training have led to dramatic improvements in few-shot learning of NLP tasks, there is limited understanding of what drives successful few-shot adaptation in datasets. In particular, given a new dataset and a pre-trained model, what properties of the dataset make it few-shot learnable, and are these properties independent of the specific adaptation techniques used? We consider an extensive set of recent few-shot learning methods and show that their performance across a large number of datasets is highly correlated, showing that few-shot hardness may be intrinsic to datasets, for a given pre-trained model. To estimate intrinsic few-shot hardness, we then propose a simple and lightweight metric called Spread that captures the intuition that few-shot learning is made possible by exploiting feature-space invariances between training and test samples. Our metric better accounts for few-shot hardness compared to existing notions of hardness and is ~8-100x faster to compute.</abstract>
<identifier type="citekey">zhao-etal-2022-measuring</identifier>
<identifier type="doi">10.18653/v1/2022.emnlp-main.262</identifier>
<location>
<url>https://aclanthology.org/2022.emnlp-main.262/</url>
</location>
<part>
<date>2022-12</date>
<extent unit="page">
<start>3955</start>
<end>3963</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T On Measuring the Intrinsic Few-Shot Hardness of Datasets
%A Zhao, Xinran
%A Murty, Shikhar
%A Manning, Christopher
%Y Goldberg, Yoav
%Y Kozareva, Zornitsa
%Y Zhang, Yue
%S Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing
%D 2022
%8 December
%I Association for Computational Linguistics
%C Abu Dhabi, United Arab Emirates
%F zhao-etal-2022-measuring
%X While advances in pre-training have led to dramatic improvements in few-shot learning of NLP tasks, there is limited understanding of what drives successful few-shot adaptation in datasets. In particular, given a new dataset and a pre-trained model, what properties of the dataset make it few-shot learnable, and are these properties independent of the specific adaptation techniques used? We consider an extensive set of recent few-shot learning methods and show that their performance across a large number of datasets is highly correlated, showing that few-shot hardness may be intrinsic to datasets, for a given pre-trained model. To estimate intrinsic few-shot hardness, we then propose a simple and lightweight metric called Spread that captures the intuition that few-shot learning is made possible by exploiting feature-space invariances between training and test samples. Our metric better accounts for few-shot hardness compared to existing notions of hardness and is ~8-100x faster to compute.
%R 10.18653/v1/2022.emnlp-main.262
%U https://aclanthology.org/2022.emnlp-main.262/
%U https://doi.org/10.18653/v1/2022.emnlp-main.262
%P 3955-3963
Markdown (Informal)
[On Measuring the Intrinsic Few-Shot Hardness of Datasets](https://aclanthology.org/2022.emnlp-main.262/) (Zhao et al., EMNLP 2022)
ACL
- Xinran Zhao, Shikhar Murty, and Christopher Manning. 2022. On Measuring the Intrinsic Few-Shot Hardness of Datasets. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 3955–3963, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.