Pre-trained Vision and Language Transformers achieve high performance on downstream tasks due to their ability to transfer representational knowledge accumulated during pretraining on substantial amounts of data. In this paper, we ask whether it is possible to compete with such models using features based on transferred (pre-trained, frozen) representations combined with a lightweight architecture. We take a multimodal guessing task as our testbed, GuessWhat?!. An ensemble of our lightweight model matches the performance of the finetuned pre-trained transformer (LXMERT). An uncertainty analysis of our ensemble shows that the lightweight transferred representations close the data uncertainty gap with LXMERT, while retaining model diversity leading to ensemble boost. We further demonstrate that LXMERT’s performance gain is due solely to its extra V&L pretraining rather than because of architectural improvements. These results argue for flexible integration of multiple features and lightweight models as a viable alternative to large, cumbersome, pre-trained models.
A Small but Informed and Diverse Model: The Case of the Multimodal GuessWhat!? Guessing Game / Greco, Claudio; Testoni, Alberto; Bernardi, Raffaella; Frank, Stella. - ELETTRONICO. - (2022), pp. 1-10. (Intervento presentato al convegno CLASP tenutosi a Gothenburg nel 15-16 September 2022).
A Small but Informed and Diverse Model: The Case of the Multimodal GuessWhat!? Guessing Game
Greco, Claudio;Testoni, Alberto;Bernardi, Raffaella;Frank, Stella
2022-01-01
Abstract
Pre-trained Vision and Language Transformers achieve high performance on downstream tasks due to their ability to transfer representational knowledge accumulated during pretraining on substantial amounts of data. In this paper, we ask whether it is possible to compete with such models using features based on transferred (pre-trained, frozen) representations combined with a lightweight architecture. We take a multimodal guessing task as our testbed, GuessWhat?!. An ensemble of our lightweight model matches the performance of the finetuned pre-trained transformer (LXMERT). An uncertainty analysis of our ensemble shows that the lightweight transferred representations close the data uncertainty gap with LXMERT, while retaining model diversity leading to ensemble boost. We further demonstrate that LXMERT’s performance gain is due solely to its extra V&L pretraining rather than because of architectural improvements. These results argue for flexible integration of multiple features and lightweight models as a viable alternative to large, cumbersome, pre-trained models.File | Dimensione | Formato | |
---|---|---|---|
2022.clasp-1.1.pdf
accesso aperto
Descrizione: paper
Tipologia:
Versione editoriale (Publisher’s layout)
Licenza:
Creative commons
Dimensione
535.72 kB
Formato
Adobe PDF
|
535.72 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione