AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Rethinking Knowledge Transfer in Image Quality Assessment: A Perceptual Preference Structure Alignment Perspective

Aobo Li, Jinjian Wu, Yongxu Liu, Jupo Ma, Weisheng Dong

CVPR 2026 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Geometric alignment and synthetic-to-real adaptation

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
select 20% preference-aligned source samples (PreSTA-S) instead of full KADID-10kSwin-B trained on KADID-10k · LIVEC cross-domainSRCC: 0.589 -> 0.744Improved
select 20% preference-aligned samples of KADID-10kSwin-B · BID cross-domainSRCC: 0.739 -> 0.833Improved
select 20% preference-aligned samples of KonIQ-10kSwin-B · TID2013 cross-domainSRCC: 0.452 -> 0.516Improved
PreSTA-S on KonIQ-10k source for CSIQ targetSwin-B · CSIQ cross-domainSRCC: 0.723 -> 0.720No clear effect
PreSTA-S within synthetic: KADID-10k to LIVE with 20% dataSwin-B · LIVE within-domainSRCC: 0.934 -> 0.933No clear effect
PreSTA-S within synthetic: KADID-10k to CSIQ with 27.9% dataSwin-B · CSIQ within-domainSRCC: 0.773 -> 0.840Improved
add all KADID-10k (cross-domain) to LIVEC trainingtarget-only Swin-B · LIVEC joint trainingSRCC: 0.883 -> 0.876Hurt
add all KonIQ-10k + SPAQ (21,198 samples) to LIVEC trainingtarget-only Swin-B · LIVEC joint trainingSRCC: 0.883 -> 0.899Improved
add only 2,793 preference-aligned samples (PreSTA-J) instead of all 21,198joint training on LIVEC · LIVEC joint trainingSRCC: 0.899 -> 0.905Improved
add 5,343 preference-aligned samples (PreSTA-J) instead of all 21,198joint training on BID · BID joint trainingSRCC: 0.883 -> 0.885Improved
use PPC to choose between KADID-10k and KonIQ-10k as sourcetransfer source selection · 12 backbone-target comparisonsSRCC: higher PPC source wins in 11 of 12Improved

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