AI Paper Atlas

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

HPSv3: Towards Wide-Spectrum Human Preference Score

Yuhang Ma, Xiaoshi Wu, Keqiang Sun, Hongsheng Li

ICCV 2025 · Image generation › Text-to-image diffusion · approach: Preference optimization for diffusion · 34 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
use Qwen2VL-7B backbone instead of Qwen2VL-2B (ablation)HPSv3 · HPDv3 test setpreference accuracy (%): 66.3 -> 76.9; gain of 10.6 ppImproved
use Qwen2VL-7B backbone instead of CLIP ViT-H/14 (ablation)HPSv3 · HPDv3 test setpreference accuracy (%): 63.5 -> 76.9; gain of 13.4 ppImproved
add uncertainty-aware ranking loss instead of ranknet loss (ablation)HPSv3 · HPDv3 test setpreference accuracy (%): 76.3 -> 76.9Improved
add uncertainty-aware ranking loss instead of ranknet loss (ablation)HPSv3 · PickScore test setpreference accuracy (%): 70.6 -> 72.8Improved
train on full HPDv3 data instead of HPDv2 data only (ablation)HPSv3 · HPDv3 test setpreference accuracy (%): 66.6 -> 76.9Improved
train on full HPDv3 data instead of ImageReward data only (ablation)HPSv3 · HPDv3 test setpreference accuracy (%): 63.4 -> 76.9Improved
train on full HPDv3 data instead of PickScore data only (ablation)HPSv3 · HPDv3 test setpreference accuracy (%): 64.9 -> 76.9Improved
run more model-wise rounds in CoHP (1 to 4) (ablation)CoHP · HPDv3 Benchmark (1000 prompts)HPSv3 score: 11.34 -> 11.69Improved
run a fifth model-wise round in CoHP (ablation)CoHP · HPDv3 Benchmark (1000 prompts)HPSv3 score: 11.69 -> 11.65No clear effect
run more sample-wise rounds in CoHP (1 to 4) (ablation)CoHP · HPDv3 Benchmark (1000 prompts)HPSv3 score: 11.59 -> 12.84Improved
use HPSv3 instead of ImageReward as CoHP preference modelCoHP · user study on 100 text-image pairswin rate: 87%Improved
use HPSv3 as reward in DanceGRPO instead of HPSv2SD1.4 with DanceGRPO · qualitativereward hackingImproved

Later papers that compared with this method

9 later accepted papers report HPSv3 as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.

1 more not listed.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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