Towards Compatible Fine-tuning for Vision-Language Model Updates

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Zhengbo, Liang, Jian, Sheng, Lijun, He, Ran, Wang, Zilei, Tan, Tieniu
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909444191813632
author Wang, Zhengbo
Liang, Jian
Sheng, Lijun
He, Ran
Wang, Zilei
Tan, Tieniu
author_facet Wang, Zhengbo
Liang, Jian
Sheng, Lijun
He, Ran
Wang, Zilei
Tan, Tieniu
contents So far, efficient fine-tuning has become a popular strategy for enhancing the capabilities of foundation models on downstream tasks by learning plug-and-play modules. However, existing methods overlook a crucial issue: if the underlying foundation model is updated, are these plug-and-play modules still effective? In this paper, we first conduct a detailed analysis of various fine-tuning methods on the CLIP in terms of their compatibility with model updates. The study reveals that many high-performing fine-tuning methods fail to be compatible with the upgraded models. To address this, we propose a novel approach, Class-conditioned Context Optimization (ContCoOp), which integrates learnable prompts with class embeddings using an attention layer before inputting them into the text encoder. Consequently, the prompts can dynamically adapt to the changes in embedding space (due to model updates), ensuring continued effectiveness. Extensive experiments over 15 datasets show that our ContCoOp achieves the highest compatibility over the baseline methods, and exhibits robust out-of-distribution generalization.
format Preprint
id arxiv_https___arxiv_org_abs_2412_20895
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Towards Compatible Fine-tuning for Vision-Language Model Updates
Wang, Zhengbo
Liang, Jian
Sheng, Lijun
He, Ran
Wang, Zilei
Tan, Tieniu
Computer Vision and Pattern Recognition
Machine Learning
So far, efficient fine-tuning has become a popular strategy for enhancing the capabilities of foundation models on downstream tasks by learning plug-and-play modules. However, existing methods overlook a crucial issue: if the underlying foundation model is updated, are these plug-and-play modules still effective? In this paper, we first conduct a detailed analysis of various fine-tuning methods on the CLIP in terms of their compatibility with model updates. The study reveals that many high-performing fine-tuning methods fail to be compatible with the upgraded models. To address this, we propose a novel approach, Class-conditioned Context Optimization (ContCoOp), which integrates learnable prompts with class embeddings using an attention layer before inputting them into the text encoder. Consequently, the prompts can dynamically adapt to the changes in embedding space (due to model updates), ensuring continued effectiveness. Extensive experiments over 15 datasets show that our ContCoOp achieves the highest compatibility over the baseline methods, and exhibits robust out-of-distribution generalization.
title Towards Compatible Fine-tuning for Vision-Language Model Updates
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2412.20895