AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

PIPBench: A Profile-Inclusive Framework for Personalized Image Generation Evaluation

Yuhang Wu, Shuxiang Zhang, WEE CHING, Chi Zhang, Miao Liu

ECCV 2026 · Image generation › Text-to-image diffusion · approach: Multi-subject personalization · 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 GPT-5 VLM condition fusion over reference imagesQwen-Image · PIPBench Real-UserElo: 1427 -> 1765Improved
add GPT-5 VLM condition fusion over reference imagesQwen-Image · PIPBench Real-UserDIS-R: 12.099 -> 22.160Improved
use two reference images instead of one in Qwen-Image-EditQwen-Image-Edit · PIPBench Real-UserElo: 1521 -> 1354Hurt
add preference conditioning via GPT-5 fusionQwen-Image · PIPBench Real-UserCLS-T: 32.677 -> 30.550Hurt
scale the fusion VLM from QwenVL2.5-7B to 70BVLM condition fusion · PIPBench Real-UserElo: 1445 -> 1531Improved
replace SDXL with Qwen-Image as generation backbone under GPT-5 fusionGPT-5 fusion · PIPBench Real-UserElo: 1394 -> 1603Improved
instruction-tune Qwen2.5-VL-7B on synthetic agent dataVLM caption enrichment · PIPBench Real-UserElo: 1492 -> 1597Improved
unfreeze the VLM in the preference compressor (ablation)preference compression · PIPBench Real-UserElo: 1318 -> 1534Improved
use profile-inclusive synthetic data instead of profile-free for VLM tuning (ablation)Qwen2.5-VL-7B · Real-Userwin rate: 47.6% -> 52.4%Improved
use profile-inclusive synthetic data instead of profile-free for VLM tuning (ablation)Qwen2.5-VL-7B · Real-UserLPIPS-R: 0.7192 -> 0.6901Improved
use profile-inclusive synthetic data instead of profile-free for VLM tuning (ablation)Qwen2.5-VL-7B · Real-UserCLS-T: 32.251 -> 31.588Hurt
add user profile conditioning to synthetic benchmark generation (ablation)benchmark construction · PIPBench synthetic dataShannon entropy: 0.9919 -> 0.5811Improved

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