Domain Generalization via Text-Anchored Information Bottleneck

2026-07-01

AI Summary

This work addresses the limited generalization of visual recognition models in novel environments due to domain-specific spurious correlations. It proposes a novel paradigm that leverages the language embedding space as a primary source of domain invariance, introducing a text-anchored information bottleneck mechanism to suppress misleading visual cues while preserving essential semantic content. Shifting the focus of domain generalization from representation learning to the design of invariance-aware supervision, the method builds upon large vision-language models to construct a training framework guided by textual semantics. Extensive experiments demonstrate state-of-the-art performance across multiple backbone architectures, confirming the effectiveness and broad applicability of text-guided supervision in enhancing model robustness and generalization.

Abstract

Visual recognition models often fail when deployed in new environments. Domain Generalization (DG) addresses this by learning representations that remain invariant to environment-specific variations. Recent approaches increasingly rely on large vision-language models, assuming that preserving their expressive visual representations improves robustness. However, we show that such visual expressiveness can instead propagate spurious cues that tie representations to the training environments, hindering invariant learning. We therefore discard visual guidance and instead treat the language embedding space as the primary source of domain invariance, naturally acting as an information bottleneck that preserves core semantics while suppressing domain-specific variations. Extensive experiments across diverse backbones exhibit state-of-the-art performance and further analyze what makes guidance effective for robust generalization. These findings shift the focus of DG from improving representations to designing supervision that enforces invariance.

Problem

• Domain Generalization
• Visual Recognition
• Spurious Correlations
• Invariant Representation
• Information Bottleneck

Innovation

• Domain Generalization
• Information Bottleneck
• Vision-Language Models
• Language Embedding
• Invariant Learning