Concept-Guided Fine-Tuning: Steering ViTs away from Spurious Correlations to Improve Robustness

TL;DR

This paper introduces a concept-guided fine-tuning method for Vision Transformers that enhances robustness to distribution shifts by aligning model relevance with automatically generated, concept-level masks, reducing reliance on spurious correlations.

Contribution

It proposes a novel, automatic, concept-level supervision framework that improves ViT robustness and interpretability without manual annotations.

Findings

Improved robustness across five out-of-distribution benchmarks.

Relevance maps better align with semantic object parts.

Concept-guided masks outperform traditional segmentation maps.

Abstract

Vision Transformers (ViTs) often degrade under distribution shifts because they rely on spurious correlations, such as background cues, rather than semantically meaningful features. Existing regularization methods, typically relying on simple foreground-background masks, which fail to capture the fine-grained semantic concepts that define an object (e.g., ``long beak'' and ``wings'' for a ``bird''). As a result, these methods provide limited robustness to distribution shifts. To address this limitation, we introduce a novel finetuning framework that steers model reasoning toward concept-level semantics. Our approach optimizes the model's internal relevance maps to align with spatially grounded concept masks. These masks are generated automatically, without manual annotation: class-relevant concepts are first proposed using an LLM-based, label-free method, and then segmented using a VLM. The finetuning objective aligns relevance with these concept regions while simultaneously suppressing focus on spurious background areas. Notably, this process requires only a minimal set of images and uses half of the dataset classes. Extensive experiments on five out-of-distribution benchmarks demonstrate that our method improves robustness across multiple ViT-based models. Furthermore, we show that the resulting relevance maps exhibit stronger alignment with semantic object parts, offering a scalable path toward more robust and interpretable vision models. Finally, we confirm that concept-guided masks provide more effective supervision for model robustness than conventional segmentation maps, supporting our central hypothesis.