OTTER: Effortless Label Distribution Adaptation of Zero-shot Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912094730846208 |
|---|---|
| author | Shin, Changho Zhao, Jitian Cromp, Sonia Vishwakarma, Harit Sala, Frederic |
| author_facet | Shin, Changho Zhao, Jitian Cromp, Sonia Vishwakarma, Harit Sala, Frederic |
| contents | Popular zero-shot models suffer due to artifacts inherited from pretraining. One particularly detrimental issue, caused by unbalanced web-scale pretraining data, is mismatched label distribution. Existing approaches that seek to repair the label distribution are not suitable in zero-shot settings, as they have mismatching requirements, such as needing access to labeled downstream task data or knowledge of the true label balance in the pretraining distribution. We sidestep these challenges and introduce a simple and lightweight approach to adjust pretrained model predictions via optimal transport. Our technique requires only an estimate of the label distribution of a downstream task. Theoretically, we characterize the improvement produced by our procedure under certain mild conditions and provide bounds on the error caused by misspecification. Empirically, we validate our method in a wide array of zero-shot image and text classification tasks, improving accuracy by 4.8% and 15.9% on average, and beating baselines like prior matching -- often by significant margins -- in 17 out of 21 datasets. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_08461 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | OTTER: Effortless Label Distribution Adaptation of Zero-shot Models Shin, Changho Zhao, Jitian Cromp, Sonia Vishwakarma, Harit Sala, Frederic Machine Learning Artificial Intelligence Popular zero-shot models suffer due to artifacts inherited from pretraining. One particularly detrimental issue, caused by unbalanced web-scale pretraining data, is mismatched label distribution. Existing approaches that seek to repair the label distribution are not suitable in zero-shot settings, as they have mismatching requirements, such as needing access to labeled downstream task data or knowledge of the true label balance in the pretraining distribution. We sidestep these challenges and introduce a simple and lightweight approach to adjust pretrained model predictions via optimal transport. Our technique requires only an estimate of the label distribution of a downstream task. Theoretically, we characterize the improvement produced by our procedure under certain mild conditions and provide bounds on the error caused by misspecification. Empirically, we validate our method in a wide array of zero-shot image and text classification tasks, improving accuracy by 4.8% and 15.9% on average, and beating baselines like prior matching -- often by significant margins -- in 17 out of 21 datasets. |
| title | OTTER: Effortless Label Distribution Adaptation of Zero-shot Models |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2404.08461 |