AI Paper Atlas

Research map / Image generation / Image restoration

Bridging Information Asymmetry: A Hierarchical Framework for Deterministic Blind Face Restoration

Zhengjian Yao, Jiakui Hu, Kaiwen Li, Hangzhou He, Xinliang Zhang, Shuang Zeng, Lei Zhu, Yanye Lu

TPAMI 2026 · Image generation › Image restoration · approach: Efficient restoration architectures · 3 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 Stage 2 fidelity-constrained preference optimization (ablation)Pref-Restore · CelebA-Test (synthetic)LPIPS: 0.3916 -> 0.3657Improved
add Stage 2 fidelity-constrained preference optimization (ablation)Pref-Restore · CelebA-Test (synthetic)FID(HQ): 13.9015 -> 23.7305Hurt
add Stage 2 fidelity-constrained preference optimization (ablation)Pref-Restore · CelebA-Test (synthetic)ArcFace Deg: 54.9791 -> 51.5394Improved
add Stage 1.2 texture-to-diffusion alignment after Stage 1.1 (ablation)Pref-Restore · CelebA-Test, staged alignment ablationArcFace Deg: 74.5286 -> 60.9876Improved
add Stage 1.2 texture-to-diffusion alignment after Stage 1.1 (ablation)Pref-Restore · CelebA-Test, staged alignment ablationLPIPS: 0.5272 -> 0.4413Improved
train semantic and texture alignment sequentially instead of jointly (ablation)Pref-Restore · CelebA-Test, staged alignment ablationArcFace Deg: 62.1582 -> 60.9876Improved
use aesthetic-only reward in Stage 2 (ablation)Pref-Restore · CelebA-Test, Stage 2 reward ablationArcFace Deg: 54.0623 -> 73.9882Hurt
add identity reward to aesthetic reward (ablation)Pref-Restore · CelebA-Test, Stage 2 reward ablationArcFace Deg: 73.9882 -> 47.1055Improved
add geometry reward to identity + LPIPS rewards (ablation)Pref-Restore · CelebA-Test, Stage 2 reward ablationArcFace Deg: 47.0899 -> 49.6654Hurt
add AR semantic branch without text (ablation)Pref-Restore · CelebA-Test, text guidanceArcFace Deg: 63.0939 -> 60.5767Improved
provide ground-truth captions at inference (ablation)Pref-Restore (Train-Caption) · CelebA-Test, text guidanceArcFace Deg: 60.9876 -> 54.3705Improved
provide captions generated from degraded input at inference (ablation)Pref-Restore (Train-Caption) · CelebA-Test, text guidanceArcFace Deg: 60.9876 -> 58.6068Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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