Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts
About
The fine-tuning paradigm in addressing long-tail learning tasks has sparked significant interest since the emergence of foundation models. Nonetheless, how fine-tuning impacts performance in long-tail learning was not explicitly quantified. In this paper, we disclose that heavy fine-tuning may even lead to non-negligible performance deterioration on tail classes, and lightweight fine-tuning is more effective. The reason is attributed to inconsistent class conditions caused by heavy fine-tuning. With the observation above, we develop a low-complexity and accurate long-tail learning algorithms LIFT with the goal of facilitating fast prediction and compact models by adaptive lightweight fine-tuning. Experiments clearly verify that both the training time and the learned parameters are significantly reduced with more accurate predictive performance compared with state-of-the-art approaches. The implementation code is available at https://github.com/shijxcs/LIFT.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Long-Tailed Image Classification | ImageNet-LT (test) | Top-1 Acc (Overall)77.8 | 220 | |
| Image Classification | ImageNet-LT (test) | Top-1 Acc (All)78.3 | 159 | |
| Image Classification | Places-LT (test) | Accuracy (Medium)53.1 | 128 | |
| Image Classification | iNaturalist 2018 (val) | -- | 116 | |
| Long-tailed Visual Recognition | ImageNet LT | Overall Accuracy82.9 | 89 | |
| Long-Tailed Image Classification | iNaturalist 2018 | Accuracy85.2 | 82 | |
| Image Classification | ImageNet-LT (val) | Top-1 Acc (Total)78.3 | 72 | |
| Image Classification | CIFAR-100 Imbalance Ratio LT-50 (test) | Accuracy90.2 | 62 | |
| Image Classification | CIFAR-100-LT Imbalance Ratio 100 (test) | Accuracy89.1 | 62 | |
| Long-Tailed Image Classification | Places-LT (test) | Accuracy51.8 | 61 |