Cross-Modal Fine-Tuning: Align then Refine

TL;DR

ORCA is a versatile cross-modal fine-tuning framework that aligns and refines pretrained models to perform well across diverse modalities and datasets, achieving state-of-the-art results.

Contribution

It introduces a novel align-then-refine workflow enabling a single pretrained model to adapt effectively to multiple modalities.

Findings

State-of-the-art results on 3 benchmarks with 60+ datasets from 12 modalities

Effective data alignment improves performance, especially in data-limited scenarios

Outperforms various hand-designed, AutoML, and task-specific methods

Abstract

Fine-tuning large-scale pretrained models has led to tremendous progress in well-studied modalities such as vision and NLP. However, similar gains have not been observed in many other modalities due to a lack of relevant pretrained models. In this work, we propose ORCA, a general cross-modal fine-tuning framework that extends the applicability of a single large-scale pretrained model to diverse modalities. ORCA adapts to a target task via an align-then-refine workflow: given the target input, ORCA first learns an embedding network that aligns the embedded feature distribution with the pretraining modality. The pretrained model is then fine-tuned on the embedded data to exploit the knowledge shared across modalities. Through extensive experiments, we show that ORCA obtains state-of-the-art results on 3 benchmarks containing over 60 datasets from 12 modalities, outperforming a wide range of hand-designed, AutoML, general-purpose, and task-specific methods. We highlight the importance of data alignment via a series of ablation studies and demonstrate ORCA's utility in data-limited regimes.