Where Does Linguistic Information Emerge in Neural Language Models? Measuring Gains and Contributions across Layers
The result's identifiers
Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F22%3A8T27XUX5" target="_blank" >RIV/00216208:11320/22:8T27XUX5 - isvavai.cz</a>
Result on the web
<a href="https://aclanthology.org/2022.coling-1.413" target="_blank" >https://aclanthology.org/2022.coling-1.413</a>
DOI - Digital Object Identifier
—
Alternative languages
Result language
angličtina
Original language name
Where Does Linguistic Information Emerge in Neural Language Models? Measuring Gains and Contributions across Layers
Original language description
Probing studies have extensively explored where in neural language models linguistic information is located. The standard approach to interpreting the results of a probing classifier is to focus on the layers whose representations give the highest performance on the probing task. We propose an alternative method that asks where the task-relevant information emerges in the model. Our framework consists of a family of metrics that explicitly model local information gain relative to the previous layer and each layer's contribution to the model's overall performance. We apply the new metrics to two pairs of syntactic probing tasks with different degrees of complexity and find that the metrics confirm the expected ordering only for one of the pairs. Our local metrics show a massive dominance of the first layers, indicating that the features that contribute the most to our probing tasks are not as high-level as global metrics suggest.
Czech name
—
Czech description
—
Classification
Type
D - Article in proceedings
CEP classification
—
OECD FORD branch
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Result continuities
Project
—
Continuities
—
Others
Publication year
2022
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Data specific for result type
Article name in the collection
Proceedings of the 29th International Conference on Computational Linguistics
ISBN
—
ISSN
2951-2093
e-ISSN
—
Number of pages
13
Pages from-to
4664-4676
Publisher name
International Committee on Computational Linguistics
Place of publication
—
Event location
Gyeongju, Republic of Korea
Event date
Jan 1, 2022
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
—