Explaining Similarity in Vision-Language Encoders with Weighted Banzhaf Interactions
This work addresses the need for better explanation methods in vision-language models, which is important for researchers and practitioners in AI interpretability, though it is incremental as it builds on existing interaction frameworks.
The paper tackled the problem of explaining similarity in vision-language encoders by proposing FIxLIP, a method based on weighted Banzhaf interactions to capture complex cross-modal interactions, and it outperformed first-order attribution methods on benchmarks like MS COCO and ImageNet-1k.
Language-image pre-training (LIP) enables the development of vision-language models capable of zero-shot classification, localization, multimodal retrieval, and semantic understanding. Various explanation methods have been proposed to visualize the importance of input image-text pairs on the model's similarity outputs. However, popular saliency maps are limited by capturing only first-order attributions, overlooking the complex cross-modal interactions intrinsic to such encoders. We introduce faithful interaction explanations of LIP models (FIxLIP) as a unified approach to decomposing the similarity in vision-language encoders. FIxLIP is rooted in game theory, where we analyze how using the weighted Banzhaf interaction index offers greater flexibility and improves computational efficiency over the Shapley interaction quantification framework. From a practical perspective, we propose how to naturally extend explanation evaluation metrics, such as the pointing game and area between the insertion/deletion curves, to second-order interaction explanations. Experiments on the MS COCO and ImageNet-1k benchmarks validate that second-order methods, such as FIxLIP, outperform first-order attribution methods. Beyond delivering high-quality explanations, we demonstrate the utility of FIxLIP in comparing different models, e.g. CLIP vs. SigLIP-2.