AI Paper Atlas

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

CaPCL: Caption-Preserved Continual Learning for Text-to-Image Retrieval

Naoya Sogi, Ren Ohkubo, Takashi Shibata, Makoto Terao, Yusuke Hosoya, Takayuki Okatani

ECCV 2026 · Vision-language and multimodal › Cross-modal learning · approach: Probabilistic embeddings for cross-modal retrieval · 0 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 caption-based regularization (Lcap)BLIP-2 continual T2I · avg over 4 ordersFM: 0.1241 (FT) -> 0.0654Improved
caption regularization vs similarity-matrix distillationBLIP-2 · avg over 4 ordersAA: 0.6387 (ModX) -> 0.6942Improved
caption regularization in hardest orderBLIP-2 · PTRBIFM: 0.2674 (FT) -> 0.0965Improved
EWC over plain fine-tuningFT · avg over 4 ordersFM: 0.1241 -> 0.1241No clear effect
similarity-matrix distillation (DKR, C2MR) over EWCBLIP-2 · avg over 4 ordersAA: 0.6460 -> 0.6300Hurt
use only Lacs (similarity distillation on auxiliary captions) (ablation)CaPCL · avg over 4 ordersHM: 0.7966 -> 0.7272Hurt
use only Lacc instead of Lcap (ablation)CaPCL · avg over 4 ordersHM: 0.7966 -> 0.7684Hurt
add Lacc and Lacs to Lcap (ablation)CaPCL · avg over 4 ordersHM: 0.7966 -> 0.7987Improved
use Qwen2.5-VL captions instead of self-generated (ablation)CaPCL · avg over 4 ordersHM: 0.7966 -> 0.7884Hurt
combine self and Qwen captions (ablation)CaPCL · avg over 4 ordersHM: 0.7966 -> 0.8001Improved
apply auxiliary-caption contrastive loss to SigLIP2SigLIP2 · avg over 4 ordersHM: 0.6377 (FT) -> 0.7129Improved
high image overlap with high query divergenceBLIP-2 fine-tuning · Flickr30K MRRforgetting: rho 0.33/0.45 -> 0.59 for joint scoreHurt

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