Testing-based Black-box Extraction of Simple Models from RNNs and Transformers

Edi Muškardin; Martin Tappler; Bernhard K. Aichernig

Testing-based Black-box Extraction of Simple Models from RNNs and Transformers

Edi Muškardin, Martin Tappler, Bernhard K. Aichernig

Publikation: Beitrag in einer Fachzeitschrift › Konferenzartikel › Begutachtung

Abstract

In this technical report, we outline the testing-based black-box method used to extract simple and interpretable models from RNNs and transformers. Our work was done in the scope of the TAYSIR competition, in which it won the first place.

Originalsprache	englisch
Seiten (von - bis)	291-294
Seitenumfang	4
Fachzeitschrift	Proceedings of Machine Learning Research
Jahrgang	217
Publikationsstatus	Veröffentlicht - 2023
Veranstaltung	16th International Conference on Grammatical Inference: ICGI 2023 - Rabat, Marokko Dauer: 10 Juli 2023 → 13 Juli 2023

ASJC Scopus subject areas

Artificial intelligence
Software
Steuerungs- und Systemtechnik
Statistik und Wahrscheinlichkeit

Andere Dateien und Links

Verknüpfung zur Publikation in Scopus

TAYSIR Competition: First place
Muskardin, Edi (Empfänger/-in), Tappler, Martin (Empfänger/-in) & Aichernig, Bernhard (Empfänger/-in), 2023
Auszeichnung: Preise / Medaillen / Ehrungen

Dieses zitieren

@article{59bdeb7ebfd2486ab92081e739ece815,

title = "Testing-based Black-box Extraction of Simple Models from RNNs and Transformers",

abstract = "In this technical report, we outline the testing-based black-box method used to extract simple and interpretable models from RNNs and transformers. Our work was done in the scope of the TAYSIR competition, in which it won the first place.",

keywords = "Active automata learning, Model extraction, RNN, Transformers",

author = "Edi Mu{\v s}kardin and Martin Tappler and Aichernig, {Bernhard K.}",

note = "Publisher Copyright: {\textcopyright} 2023 E. Mu{\v s}kardin, M. Tappler & B.K. Aichernig.; 16th International Conference on Grammatical Inference : ICGI 2023, ICGI 2023 ; Conference date: 10-07-2023 Through 13-07-2023",

year = "2023",

language = "English",

volume = "217",

pages = "291--294",

journal = "Proceedings of Machine Learning Research",

issn = "2640-3498",

publisher = "ML Research Press",

}

TY - JOUR

T1 - Testing-based Black-box Extraction of Simple Models from RNNs and Transformers

AU - Muškardin, Edi

AU - Tappler, Martin

AU - Aichernig, Bernhard K.

PY - 2023

Y1 - 2023

N2 - In this technical report, we outline the testing-based black-box method used to extract simple and interpretable models from RNNs and transformers. Our work was done in the scope of the TAYSIR competition, in which it won the first place.

AB - In this technical report, we outline the testing-based black-box method used to extract simple and interpretable models from RNNs and transformers. Our work was done in the scope of the TAYSIR competition, in which it won the first place.

KW - Active automata learning

KW - Model extraction

KW - RNN

KW - Transformers

UR - http://www.scopus.com/inward/record.url?scp=85173564104&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:85173564104

SN - 2640-3498

VL - 217

SP - 291

EP - 294

JO - Proceedings of Machine Learning Research

JF - Proceedings of Machine Learning Research

T2 - 16th International Conference on Grammatical Inference

Y2 - 10 July 2023 through 13 July 2023

ER -

Testing-based Black-box Extraction of Simple Models from RNNs and Transformers

Abstract

ASJC Scopus subject areas

Andere Dateien und Links

Auszeichnungen

TAYSIR Competition: First place

Dieses zitieren