CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs

Lukas Thoma; Ivonne Weyers; Erion Çano; Stefan Schweter; Jutta L. Mueller; Benjamin Roth

doi:10.18653/v1/2023.conll-babylm.15

CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs

Lukas Thoma
, Ivonne Weyers
, Erion Çano
, Stefan Schweter
, Jutta L. Mueller
, Benjamin Roth

Veröffentlichungen: Beitrag in Buch › Beitrag in Konferenzband › Peer Reviewed

Originalsprache	Englisch
Titel	The BabyLM Challenge at the 27th Conference on Computational Natural Language Learning
Untertitel	Proceedings of the BabyLM Challenge
Herausgeber*innen	Alex Warstadt, Aaron Mueller, Leshem Choshen, Ethan Wilcox, Chengxu Zhuang, Juan Ciro, Rafael Mosquera, Bhargavi Paranjabe, Adina Williams, Tal Linzen, Ryan Cotterell
Erscheinungsort	Stroudsburg
Verlag	Association for Computational Linguistics (ACL)
Seiten	180-185
Seitenumfang	6
ISBN (Print)	978-1-952148-02-6
DOIs	https://doi.org/10.18653/v1/2023.conll-babylm.15 https://doi.org/10.18653/v1/2023.conll-babylm.15
Publikationsstatus	Veröffentlicht - 2023
Veranstaltung	BabyLM Challenge at the 27th Conference on Computational Natural Language Learning, CoNLL 2023 - Singapore, Singapur Dauer: 6 Dez. 2023 → 7 Dez. 2023

Konferenz

Konferenz	BabyLM Challenge at the 27th Conference on Computational Natural Language Learning, CoNLL 2023
Land/Gebiet	Singapur
Ort	Singapore
Zeitraum	6/12/23 → 7/12/23

ÖFOS 2012

102001 Artificial Intelligence
602011 Computerlinguistik

Zugriff auf Dokument

10.18653/v1/2023.conll-babylm.15Lizenz: CC BY-SA 3.0
10.18653/v1/2023.conll-babylm.15Lizenz: CC BY 4.0

https://aclanthology.org/2023.conll-babylm.15/

Andere Dateien und Links

Verknüpfung zur Publikation in Scopus

1 Abgeschlossen

Cognitive Plausibility of Deep Learning Language Models
Thoma, L. (Projektleiter*in), Blevins, T. (Co-Projektleiter*in), Schoenegger, L. (Projektmitarbeiter*in), Weyers, I. (Projektleiter*in) & Çano, E. (Assoziiertes Projektmitglied)
1/08/22 → 31/01/25
Projekt: Forschungsförderung

CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs
Thoma, L. (Vortragende*r) & Roth, B. (Vortragende*r)
2023
Aktivität: Vorträge › Posterpräsentation › Science to Science

Zitationsweisen

Thoma, L., Weyers, I., Çano, E., Schweter, S., Mueller, J. L., & Roth, B. (2023). CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs. in A. Warstadt, A. Mueller, L. Choshen, E. Wilcox, C. Zhuang, J. Ciro, R. Mosquera, B. Paranjabe, A. Williams, T. Linzen, & R. Cotterell (Hrsg.), The BabyLM Challenge at the 27th Conference on Computational Natural Language Learning: Proceedings of the BabyLM Challenge (S. 180-185). Association for Computational Linguistics (ACL). https://doi.org/10.18653/v1/2023.conll-babylm.15, https://doi.org/10.18653/v1/2023.conll-babylm.15

Thoma, Lukas ; Weyers, Ivonne ; Çano, Erion et al. / CogMemLM : Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs. The BabyLM Challenge at the 27th Conference on Computational Natural Language Learning: Proceedings of the BabyLM Challenge. Hrsg. / Alex Warstadt ; Aaron Mueller ; Leshem Choshen ; Ethan Wilcox ; Chengxu Zhuang ; Juan Ciro ; Rafael Mosquera ; Bhargavi Paranjabe ; Adina Williams ; Tal Linzen ; Ryan Cotterell. Stroudsburg : Association for Computational Linguistics (ACL), 2023. S. 180-185

@inproceedings{40a945d2f6574e7e9ba1a26608643d47,

title = "CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs",

author = "Lukas Thoma and Ivonne Weyers and Erion {\c C}ano and Stefan Schweter and Mueller, \{Jutta L.\} and Benjamin Roth",

note = "Funding Information: The present research was funded by the Go!Digital 3.0 grant program of the Austrian Academy of Sciences (GD3.0 2021-18 CogML). We thank our intern student C{\'e}lestin Eve for assisting us in this project. Furthermore, we would like to thank Google{\textquoteright}s TPU Research Cloud (TRC) program for giving us access to TPUs that were used for training our BabyLM models. We would also like to thank Hugging Face for providing the ability to host and perform inferencing of our models on the Hugging Face Model Hub.; BabyLM Challenge at the 27th Conference on Computational Natural Language Learning, CoNLL 2023 ; Conference date: 06-12-2023 Through 07-12-2023",

year = "2023",

doi = "10.18653/v1/2023.conll-babylm.15",

language = "English",

isbn = "978-1-952148-02-6",

pages = "180--185",

editor = "Alex Warstadt and Aaron Mueller and Leshem Choshen and Ethan Wilcox and Chengxu Zhuang and Juan Ciro and Rafael Mosquera and Bhargavi Paranjabe and Adina Williams and Tal Linzen and Ryan Cotterell",

booktitle = "The BabyLM Challenge at the 27th Conference on Computational Natural Language Learning",

publisher = "Association for Computational Linguistics (ACL)",

}

Thoma, L, Weyers, I, Çano, E, Schweter, S, Mueller, JL & Roth, B 2023, CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs. in A Warstadt, A Mueller, L Choshen, E Wilcox, C Zhuang, J Ciro, R Mosquera, B Paranjabe, A Williams, T Linzen & R Cotterell (Hrsg.), The BabyLM Challenge at the 27th Conference on Computational Natural Language Learning: Proceedings of the BabyLM Challenge. Association for Computational Linguistics (ACL), Stroudsburg, S. 180-185, BabyLM Challenge at the 27th Conference on Computational Natural Language Learning, CoNLL 2023, Singapore, Singapur, 6/12/23. https://doi.org/10.18653/v1/2023.conll-babylm.15, https://doi.org/10.18653/v1/2023.conll-babylm.15

CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs. / Thoma, Lukas; Weyers, Ivonne; Çano, Erion et al.
The BabyLM Challenge at the 27th Conference on Computational Natural Language Learning: Proceedings of the BabyLM Challenge. Hrsg. / Alex Warstadt; Aaron Mueller; Leshem Choshen; Ethan Wilcox; Chengxu Zhuang; Juan Ciro; Rafael Mosquera; Bhargavi Paranjabe; Adina Williams; Tal Linzen; Ryan Cotterell. Stroudsburg: Association for Computational Linguistics (ACL), 2023. S. 180-185.

Veröffentlichungen: Beitrag in Buch › Beitrag in Konferenzband › Peer Reviewed

TY - GEN

T1 - CogMemLM

T2 - BabyLM Challenge at the 27th Conference on Computational Natural Language Learning, CoNLL 2023

AU - Thoma, Lukas

AU - Weyers, Ivonne

AU - Çano, Erion

AU - Schweter, Stefan

AU - Mueller, Jutta L.

AU - Roth, Benjamin

N1 - Funding Information: The present research was funded by the Go!Digital 3.0 grant program of the Austrian Academy of Sciences (GD3.0 2021-18 CogML). We thank our intern student Célestin Eve for assisting us in this project. Furthermore, we would like to thank Google’s TPU Research Cloud (TRC) program for giving us access to TPUs that were used for training our BabyLM models. We would also like to thank Hugging Face for providing the ability to host and perform inferencing of our models on the Hugging Face Model Hub.

PY - 2023

Y1 - 2023

UR - https://www.scopus.com/pages/publications/85185342536

U2 - 10.18653/v1/2023.conll-babylm.15

DO - 10.18653/v1/2023.conll-babylm.15

M3 - Contribution to proceedings

AN - SCOPUS:85185342536

SN - 978-1-952148-02-6

SP - 180

EP - 185

BT - The BabyLM Challenge at the 27th Conference on Computational Natural Language Learning

A2 - Warstadt, Alex

A2 - Mueller, Aaron

A2 - Choshen, Leshem

A2 - Wilcox, Ethan

A2 - Zhuang, Chengxu

A2 - Ciro, Juan

A2 - Mosquera, Rafael

A2 - Paranjabe, Bhargavi

A2 - Williams, Adina

A2 - Linzen, Tal

A2 - Cotterell, Ryan

PB - Association for Computational Linguistics (ACL)

CY - Stroudsburg

Y2 - 6 December 2023 through 7 December 2023

ER -

Thoma L, Weyers I, Çano E, Schweter S, Mueller JL , Roth B. CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs. in Warstadt A, Mueller A, Choshen L, Wilcox E, Zhuang C, Ciro J, Mosquera R, Paranjabe B, Williams A, Linzen T, Cotterell R, Hrsg., The BabyLM Challenge at the 27th Conference on Computational Natural Language Learning: Proceedings of the BabyLM Challenge. Stroudsburg: Association for Computational Linguistics (ACL). 2023. S. 180-185 doi: 10.18653/v1/2023.conll-babylm.15, 10.18653/v1/2023.conll-babylm.15

CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs

Konferenz

ÖFOS 2012

Zugriff auf Dokument

Andere Dateien und Links

Projekte

Cognitive Plausibility of Deep Learning Language Models

Aktivitäten

CogMemLM: Human-Like Memory Mechanisms Improve Performance and Cognitive Plausibility of LLMs

Zitationsweisen