Stationary representations: Optimally approximating compatibility and implications for improved model replacements

N Biondi, F Pernici, S Ricci… - Proceedings of the …, 2024 - openaccess.thecvf.com
Learning compatible representations enables the interchangeable use of semantic features
as models are updated over time. This is particularly relevant in search and retrieval systems …

Towards cross-modal backward-compatible representation learning for vision-language models

YK Jang, S Lim - arxiv preprint arxiv:2405.14715, 2024 - arxiv.org
Modern retrieval systems often struggle with upgrading to new and more powerful models
due to the incompatibility of embeddings between the old and new models. This …

Asymmetric image retrieval with cross model compatible ensembles

A Shoshan, O Linial, N Bhonker… - Proceedings of the …, 2024 - openaccess.thecvf.com
The asymmetrical retrieval setting is a well suited solution for resource constrained
applications such as face recognition and image retrieval. In this setting, a large model is …

Backward-Compatible Aligned Representations via an Orthogonal Transformation Layer

S Ricci, N Biondi, F Pernici, A Del Bimbo - arxiv preprint arxiv:2408.08793, 2024 - arxiv.org
Visual retrieval systems face significant challenges when updating models with improved
representations due to misalignment between the old and new representations. The costly …

Asymmetric Image Retrieval with Cross Model Compatible Ensembles

O Linial, A Shoshan, N Bhonker, E Hirsch… - arxiv preprint arxiv …, 2023 - arxiv.org
The asymmetrical retrieval setting is a well suited solution for resource constrained
applications such as face recognition and image retrieval. In this setting, a large model is …