[NeurIPS 2023] Generalized Logit Adjustment: Calibrating Fine-tuned Models by Removing Label Bias in Foundation Models
Foundation models like CLIP allow zero-shot transfer on various tasks without additional training data. Yet, the zero-shot performance is less competitive than a fully supervised one. Thus, to enhance the performance, fine-tuning and ensembling are also commonly adopted to better fit the downstream tasks. However, we argue that such prior work has overlooked the inherent biases in foundation models. Due to the highly imbalanced Web-scale training set, these foundation models are inevitably skewed toward frequent semantics, and thus the subsequent fine-tuning or ensembling is still biased. In this study, we systematically examine the biases in foundation models and demonstrate the efficacy of our proposed Generalized Logit Adjustment (GLA) method. Note that bias estimation in foundation models is challenging, as most pre-train data cannot be explicitly accessed like in traditional long-tailed classification tasks. To this end, GLA has an optimization-based bias estimation approach for debiasing foundation models. As our work resolves a fundamental flaw in the pre-training, the proposed GLA demonstrates significant improvements across a diverse range of tasks: it achieves 1.5 pp accuracy gains on ImageNet, an large average improvement (1.4-4.6 pp) on 11 few-shot datasets, 2.4 pp gains on long-tailed classification.
- 19 Oct 2023. Due to my busy schedule, I haven't had the time to organize all the code. For now, I've uploaded the code from the supplementary materials submitted for NeurIPS. The code in code.zip primarily contains the many-shot experiment results for ImageNet1K. I will update the Few-shot and Longtail code in the future. Thank you for your patience
- 18 Nov 2023. Thanks for your patience again. I just upload the few-shot training and evalution code in GLA.FS.public.zip. The code also includes the log(q) estimating process. I will update the Long-tail code and log(q) estimtation code of many-shot before the commencement of NeurIPS 2023.
- 21 Apr 2024. I forgot to upload
Dassl'' package, please install theDassl.ProGrad.pytorch'' first then run the ``GLA.FS.public''. `