TOAST: Transfer Learning via Top-Down Attention Steering

Abstract

Transfer learning involves adapting a pre-trained model to novel downstream tasks. In this work, we empirically observe that current transfer learning methods often fail to focus on task-relevant features, potentially degrading transfer learning performance. We then explore refocusing model attention to improve transfer learning. We introduce Top-Down Attention Steering (TOAST), a novel transfer learning algorithm that keeps the pre-trained backbone frozen, selects task-relevant features in the output, and feeds those features back to the model to steer the attention to the task-specific features. By refocusing the attention only, TOAST achieves state-of-the-art results on a number of transfer learning benchmarks, while having a small number of tunable parameters. Compared to fully fine-tuning, LoRA, and prompt tuning, TOAST substantially improves performance across a range of fine-grained visual classification datasets (e.g., 82.6% $\rightarrow$ 86.2% on FGVC). TOAST also outperforms the fully fine-tuned Alpaca and Vicuna models on instruction-following language generation.