Towards Calibrated Robust Fine-Tuning of Vision-Language Models

TL;DR

This paper introduces a robust fine-tuning method for vision-language models that enhances out-of-distribution accuracy and confidence calibration by leveraging a novel theoretical insight and a constrained contrastive loss.

Contribution

It presents a new framework that improves OOD performance and calibration by enforcing a larger smallest singular value during fine-tuning, guided by self-distillation.

Findings

Improved OOD accuracy on ImageNet benchmarks.

Enhanced confidence calibration in vision-language models.

Theoretical bounds linking calibration errors and data covariance.

Abstract

Improving out-of-distribution (OOD) generalization during in-distribution (ID) adaptation is a primary goal of robust fine-tuning of zero-shot models beyond naive fine-tuning. However, despite decent OOD generalization performance from recent robust fine-tuning methods, confidence calibration for reliable model output has not been fully addressed. This work proposes a robust fine-tuning method that improves both OOD accuracy and confidence calibration simultaneously in vision language models. Firstly, we show that both OOD classification and OOD calibration errors have a shared upper bound consisting of two terms of ID data: 1) ID calibration error and 2) the smallest singular value of the ID input covariance matrix. Based on this insight, we design a novel framework that conducts fine-tuning with a constrained multimodal contrastive loss enforcing a larger smallest singular value, which is further guided by the self-distillation of a moving-averaged model to achieve calibrated prediction as well. Starting from empirical evidence supporting our theoretical statements, we provide extensive experimental results on ImageNet distribution shift benchmarks that demonstrate the effectiveness of our theorem and its practical implementation.