Test-Time Spectrum-Aware Latent Steering for Zero-Shot Generalization in Vision-Language Models

TL;DR

This paper introduces Spectrum-Aware Test-Time Steering (STS), a lightweight, inference-only adaptation method for Vision-Language Models that improves zero-shot generalization under domain shifts by steering latent representations without modifying the core model.

Contribution

STS is a novel, spectral subspace-based test-time adaptation framework that operates entirely in the latent space, requiring minimal parameters and no backpropagation.

Findings

Outperforms or matches state-of-the-art test-time adaptation methods.

Achieves up to 8x faster inference speed.

Uses 12x less memory than prompt tuning.

Abstract

Vision-Language Models (VLMs) excel at zero-shot inference but often degrade under test-time domain shifts. For this reason, episodic test-time adaptation strategies have recently emerged as powerful techniques for adapting VLMs to a single unlabeled image. However, existing adaptation strategies, such as test-time prompt tuning, typically require backpropagating through large encoder weights or altering core model components. In this work, we introduce Spectrum-Aware Test-Time Steering (STS), a lightweight adaptation framework that extracts a spectral subspace from the textual embeddings to define principal semantic directions and learns to steer latent representations in a spectrum-aware manner by adapting a small number of per-sample shift parameters to minimize entropy across augmented views. STS operates entirely at inference in the latent space, without backpropagation through or modification of the frozen encoders. Building on standard evaluation protocols, our comprehensive experiments demonstrate that STS largely surpasses or compares favorably against state-of-the-art test-time adaptation methods, while introducing only a handful of additional parameters and achieving inference speeds up to 8x faster with a 12x smaller memory footprint than conventional test-time prompt tuning. The code is available at https://github.com/kdafnis/STS.