Bayesian Principles Improve Prompt Learning In Vision-Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kim, Mingyu, Ko, Jongwoo, Park, Mijung
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910915055583232
author Kim, Mingyu
Ko, Jongwoo
Park, Mijung
author_facet Kim, Mingyu
Ko, Jongwoo
Park, Mijung
contents Prompt learning is a popular fine-tuning method for vision-language models due to its efficiency. It requires a small number of additional learnable parameters while significantly enhancing performance on target tasks. However, most existing methods suffer from overfitting to fine-tuning data, yielding poor generalizability. To address this, we propose a new training objective function based on a Bayesian learning principle to balance adaptability and generalizability. We derive a prior over the logits, where the mean function is parameterized by the pre-trained model, while the posterior corresponds to the fine-tuned model. This objective establishes a balance by allowing the fine-tuned model to adapt to downstream tasks while remaining close to the pre-trained model.
format Preprint
id arxiv_https___arxiv_org_abs_2504_14123
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Bayesian Principles Improve Prompt Learning In Vision-Language Models
Kim, Mingyu
Ko, Jongwoo
Park, Mijung
Artificial Intelligence
Computation and Language
Computer Vision and Pattern Recognition
Prompt learning is a popular fine-tuning method for vision-language models due to its efficiency. It requires a small number of additional learnable parameters while significantly enhancing performance on target tasks. However, most existing methods suffer from overfitting to fine-tuning data, yielding poor generalizability. To address this, we propose a new training objective function based on a Bayesian learning principle to balance adaptability and generalizability. We derive a prior over the logits, where the mean function is parameterized by the pre-trained model, while the posterior corresponds to the fine-tuned model. This objective establishes a balance by allowing the fine-tuned model to adapt to downstream tasks while remaining close to the pre-trained model.
title Bayesian Principles Improve Prompt Learning In Vision-Language Models
topic Artificial Intelligence
Computation and Language
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2504.14123