What explains the success of cross-modal fine-tuning with ORCA?

Paloma Garcia De Herreros, Vagrant Gautam, Philipp Slusallek, Dietrich Klakow, Marius Mosbach


Abstract
ORCA (Shen et al., 2023) is a recent technique for cross-modal fine-tuning, i.e., applying pre-trained transformer models to modalities beyond their training data. The technique consists primarily of training an embedder and fine-tuning the embedder and model. Despite its high performance on a variety of downstream tasks, we do not understand precisely how each of these components contribute to ORCA’s success. Therefore, we run a series of ablations and find that embedder training does not help 2D tasks at all, contrary to what the original paper posits. In 1D tasks, some amount of embedder training is necessary but more is not better. In 4 out of 6 datasets we experiment with, it is model fine-tuning that makes the biggest difference. Through our ablations and baselines, we contribute a better understanding of the individual components of ORCA.
Anthology ID:
2024.insights-1.2
Volume:
Proceedings of the Fifth Workshop on Insights from Negative Results in NLP
Month:
June
Year:
2024
Address:
Mexico City, Mexico
Editors:
Shabnam Tafreshi, Arjun Akula, João Sedoc, Aleksandr Drozd, Anna Rogers, Anna Rumshisky
Venues:
insights | WS
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
8–16
Language:
URL:
https://aclanthology.org/2024.insights-1.2
DOI:
10.18653/v1/2024.insights-1.2
Bibkey:
Cite (ACL):
Paloma Garcia De Herreros, Vagrant Gautam, Philipp Slusallek, Dietrich Klakow, and Marius Mosbach. 2024. What explains the success of cross-modal fine-tuning with ORCA?. In Proceedings of the Fifth Workshop on Insights from Negative Results in NLP, pages 8–16, Mexico City, Mexico. Association for Computational Linguistics.
Cite (Informal):
What explains the success of cross-modal fine-tuning with ORCA? (Garcia De Herreros et al., insights-WS 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.insights-1.2.pdf