TU Darmstadt / ULB / TUbiblio

IGLUE: A Benchmark for Transfer Learning across Modalities, Tasks, and Languages

Bugliarello, Emanuele ; Liu, Fangyu ; Pfeiffer, Jonas ; Reddy, Siva ; Elliott, Desmond ; Ponti, Edoardo M. ; Vulić, Ivan
Hrsg.: Chaudhuri, Kamalika ; Jegelka, Stefanie ; Song, Le ; Szepesvari, Csaba ; Niu, Gang ; Sabato, Sivan ; University of Copenhagen, Mila – Quebec Artificial Intelligence Institute, University of Cambridge, TU Darmstadt, New York University, McGill University (2022)
IGLUE: A Benchmark for Transfer Learning across Modalities, Tasks, and Languages.
The 39th International Conference on Machine Learning. Baltimore, Maryland USA (17.-23.07.2022)
Konferenzveröffentlichung, Bibliographie

Kurzbeschreibung (Abstract)

Reliable evaluation benchmarks designed for replicability and comprehensiveness have driven progress in machine learning. Due to the lack of a multilingual benchmark, however, vision-and-language research has mostly focused on English language tasks. To fill this gap, we introduce the Image-Grounded Language Understanding Evaluation benchmark. IGLUE brings together — by both aggregating pre-existing datasets and creating new ones — visual question answering, cross-modal retrieval, grounded reasoning, and grounded entailment tasks across 20 diverse languages. Our benchmark enables the evaluation of multilingual multimodal models for transfer learning, not only in a zero-shot setting, but also in newly defined few-shot learning setups. Based on the evaluation of the available state-of-the-art models, we find that translate-test transfer is superior to zero-shot transfer and that few-shot learning is hard to harness for many tasks. Moreover, downstream performance is partially explained by the amount of available unlabelled textual data for pretraining, and only weakly by the typological distance of target – source languages. We hope to encourage future research efforts in this area by releasing the benchmark to the community.

Typ des Eintrags: Konferenzveröffentlichung
Erschienen: 2022
Herausgeber: Chaudhuri, Kamalika ; Jegelka, Stefanie ; Song, Le ; Szepesvari, Csaba ; Niu, Gang ; Sabato, Sivan
Autor(en): Bugliarello, Emanuele ; Liu, Fangyu ; Pfeiffer, Jonas ; Reddy, Siva ; Elliott, Desmond ; Ponti, Edoardo M. ; Vulić, Ivan
Art des Eintrags: Bibliographie
Titel: IGLUE: A Benchmark for Transfer Learning across Modalities, Tasks, and Languages
Sprache: Englisch
Publikationsjahr: 6 September 2022
Ort: Baltimore, Maryland, USA
Verlag: PMLR
Buchtitel: Proceedings of the 39th International Conference on Machine Learning
Reihe: Proceedings of Machine Learning Research
Band einer Reihe: 162
Veranstaltungstitel: The 39th International Conference on Machine Learning
Veranstaltungsort: Baltimore, Maryland USA
Veranstaltungsdatum: 17.-23.07.2022
URL / URN: https://proceedings.mlr.press/v162/bugliarello22a.html
Kurzbeschreibung (Abstract):

Reliable evaluation benchmarks designed for replicability and comprehensiveness have driven progress in machine learning. Due to the lack of a multilingual benchmark, however, vision-and-language research has mostly focused on English language tasks. To fill this gap, we introduce the Image-Grounded Language Understanding Evaluation benchmark. IGLUE brings together — by both aggregating pre-existing datasets and creating new ones — visual question answering, cross-modal retrieval, grounded reasoning, and grounded entailment tasks across 20 diverse languages. Our benchmark enables the evaluation of multilingual multimodal models for transfer learning, not only in a zero-shot setting, but also in newly defined few-shot learning setups. Based on the evaluation of the available state-of-the-art models, we find that translate-test transfer is superior to zero-shot transfer and that few-shot learning is hard to harness for many tasks. Moreover, downstream performance is partially explained by the amount of available unlabelled textual data for pretraining, and only weakly by the typological distance of target – source languages. We hope to encourage future research efforts in this area by releasing the benchmark to the community.

Freie Schlagworte: UKP_p_emergencity, emergenCITY_IN
Fachbereich(e)/-gebiet(e): 20 Fachbereich Informatik
20 Fachbereich Informatik > Ubiquitäre Wissensverarbeitung
LOEWE
LOEWE > LOEWE-Zentren
LOEWE > LOEWE-Zentren > emergenCITY
TU-Projekte: HMWK|III L6-519/03/05.001-(0016)|emergenCity TP Bock
Hinterlegungsdatum: 07 Sep 2022 09:29
Letzte Änderung: 16 Sep 2022 12:38
PPN: 499482603
Export:
Suche nach Titel in: TUfind oder in Google
Frage zum Eintrag Frage zum Eintrag

Optionen (nur für Redakteure)
Redaktionelle Details anzeigen Redaktionelle Details anzeigen