aiqbench analysis
This research addresses the scalability bottleneck of updating deployed vision-language models by introducing spherical interpolation to bridge incompatible embedding spaces without costly re-indexing. By leveraging geometric properties within normalized representation spaces, the method recovers backward compatibility and improves retrieval performance over standard orthogonal alignment techniques. The findings offer a practical pathway for maintaining efficient cross-modal search systems as AI models evolve rapidly.
Spherical Interpolation for Backward-Compatible Multimodal Representations
arXiv:2609.39836v1 Announce Type: cross Abstract: Contrastive vision-language models map visual and textual representations into a shared normalized embedding space, making cosine similarity the natural metric for cross-modal retrieval. A practical challenge arises during model upgrades: independently trained models generally produce incompatible representation spaces, so replacing a deployed model typically requires recomputing embeddings for the entire gallery, which is prohibitively expensive at scale. Orthogonal post-hoc alignment can partially mitigate this problem by mapping new-model queries into the old-model gallery space. However, because independently trained models can differ in fine-grained representation structure, the orthogonal alignment remains approximate, leaving a residual angular discrepancy between the old-model query and the aligned new-model query. We study whether interpolation along the spherical geodesic between these two normalized query representations can improve retrieval without re-indexing the gallery. We characterize when this path contains an interior query direction closer to an idealized retrieval-optimal direction than either endpoint, and connect this characterization to Recall@$K$ through a local margin-based certification result. Experiments across multiple benchmarks and model families show that post-alignment spherical interpolation improves over orthogonal alignment alone, recovering backward-compatibility in most evaluated settings. Consistent with our geometric characterization, per-query oracle analysis shows that retrieval-favorable interior points occur frequently in practice. Code is available at https://github.com/miccunifi/SLERP_backward_compatibility .
Abstract:Contrastive vision-language models map visual and textual representations into a shared normalized embedding space, making cosine similarity the natural metric for cross-modal retrieval. A practical challenge arises during model upgrades: independently trained models generally produce incompatible representation spaces, so replacing a deployed model typically requires recomputing embeddings for the entire gallery, which is prohibitively expensive at scale. Orthogonal post-hoc alignment can partially mitigate this problem by mapping new-model queries into the old-model gallery space. However, because independently trained models can differ in fine-grained representation structure, the orthogonal alignment remains approximate, leaving a residual angular discrepancy between the old-model query and the aligned new-model query. We study whether interpolation along the spherical geodesic between these two normalized query representations can improve retrieval without re-indexing the gallery. We characterize when this path contains an interior query direction closer to an idealized retrieval-optimal direction than either endpoint, and connect this characterization to Recall@$K$ through a local margin-based certification result. Experiments across multiple benchmarks and model families show that post-alignment spherical interpolation improves over orthogonal alignment alone, recovering backward-compatibility in most evaluated settings. Consistent with our geometric characterization, per-query oracle analysis shows that retrieval-favorable interior points occur frequently in practice. Code is available at this https URL .
Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
arXiv:2609.39836 [cs.CV]
(or arXiv:2609.39836v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.39836 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 304 words
Related coverage
A Voxel-Spacing-Aware Extension of PyRadiomics for Anisotropic Texture Analysis
arXiv · Sep 16, 2026
The Entangled Dimensions of Decision Theory
LessWrong · Jul 30, 2026
Exploring self-distilled reasoning for supervised fine-tuning with Amazon Nova
AWS · Jul 21, 2026
Grounding Functional Similarity by Invariance-Aware Model Stitching
arXiv · Jun 3, 2026