AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Multi-modal Knowledge Preserving Adapter for Embedding Backward Compatibility

Jaeseok Byun, Gukyeong Kwon, Han-Kai Hsu, MEHER GITIKA KARUMURI, Zhikang Zhang, Hao Yang, Davide Modolo

ECCV 2026 · Vision-language and multimodal › Cross-modal learning · approach: Probabilistic embeddings for cross-modal retrieval · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add multi-level preservation loss plus focal re-weighting to the base adapterL2 base adapter · MMEB, CLIP->UniME-7BPrecision@1: 20.03 -> 49.56Improved
add MKP-Adapterbase adapter · COCO/Flickr30K/Urban1K, CLIP->UniMEI2T Recall@1: 63.35 -> 77.76Improved
use L2-based adaptersbase/EC-style · T2I retrievalRecall@1: E5-V old 72.41 vs Base 72.11Hurt
use contrastive base adapterCL-based adapter · I2T retrieval, CLIP->GMERecall@1: 69.80 -> 64.59Hurt
add distance-wise preservation loss (ablation)MKP-Adapter · I2T avg, CLIP->UniMERecall@1: 71.74 -> 77.76Improved
add angle-wise preservation loss (ablation)MKP-Adapter · MMEB, CLIP->UniMEPrecision@1: 41.55 -> 49.56Improved
add point-wise preservation loss (ablation)MKP-Adapter · MMEB, CLIP->UniMEPrecision@1: 47.43 -> 49.56Improved
add point-wise preservation loss (ablation)MKP-Adapter · T2I avg, CLIP->UniMERecall@1: 63.22 (w/o) vs 63.07No clear effect
add focal re-weighting (ablation)MKP-Adapter · MMEB, CLIP->UniMEPrecision@1: 46.61 -> 49.56Improved
use a single shared adapter instead of separate per-modality adapters (ablation)MKP-Adapter · I2T/T2I, E5-V->UniMERecall@1: base L2 72.93 vs separate L2 72.30Improved
use L2 instead of L1 base loss (ablation)base adapter · I2T avg, E5-V->UniMERecall@1: 72.03 -> 72.93Improved
increase adapter size or training data from 558k to 4M pairs (ablation)MKP-Adapter · I2T/T2I, E5-V->UniMERecall@1: 82.03 -> 82.54 (4M pairs)No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy