AI Paper Atlas

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

Bridging the Perceptual Gap: Residual-Enhanced Downscaling and Manifold-Aware Perception Alignment Adaptation for NR-IQA

Yu Li, Zhengran Shen, Yachun Mi, Puchao Zhou, Shaohui Liu

ICML 2026 · Vision-language and multimodal › Cross-modal learning · approach: Contrastive vision-language alignment · 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 RPD residual-enhanced downsampling to CMPA (ablation)CMPA · KonIQ-10kSRCC: 0.935 -> 0.938Improved
add RPD residual-enhanced downsampling to CMPA (ablation)CMPA · KADID-10kSRCC: 0.937 -> 0.938Improved
add PAI cross-modal interaction on top of PFE (ablation)CMPA · KADID-10kSRCC: 0.936 -> 0.938Improved
add PFE shared MLP to dual text and image tuning (ablation)CMPA · KADID-10kSRCC: 0.928 -> 0.936Improved
tune image encoder adapter in addition to text (ablation)CMPA · KADID-10kSRCC: 0.792 -> 0.928Improved
tune text and image independently vs image only (ablation)CMPA · KADID-10kSRCC: 0.928 -> 0.928No clear effect
use low-dimensional PFE instead of high-dimensional (ablation)CMPA PFE · KADID-10kSRCC: 0.912 -> 0.936Improved
add residual information to bicubic downsampling (ablation)RPD · TID2013SRCC: 0.879 -> 0.870Hurt
add perceptual weight on top of JND weight map (ablation)RPD · TID2013SRCC: 0.883 -> 0.899Improved
add RPD preprocessing to HyperIQAHyperIQA · LIVECSRCC: 0.859 -> 0.868Improved
add RPD preprocessing to MANIQAMANIQA · LIVECSRCC: 0.868 -> 0.870Improved
use ViT-L-14 backbone instead of ViT-B-16CMPA · KADID-10kSRCC: 0.934 -> 0.951Improved

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