List your company
ResearchFormerly WizWand

Human Preference Evaluation on HPD v2 (test)

85.4Preference AccuracyHPSv3 - 7B
64.18469.69275.280.708May 23, 2024Sep 22, 2024Jan 22, 2025May 24, 2025Sep 23, 2025Jan 23, 2026May 25, 2026
Updated 5mo ago

Evaluation results

MethodLinks
HPSv3 - 7B2026.0585.4
HPSv3Backbone=Qwen2VL-7B2026.0285.36
DiNa-LRMBackbone=SD3.5-M-2B, Multi-noise inference scaling=true2026.0284.31
MPSTrained on=MHP datasets2024.0583.5
MPS2026.0583.5
HPS v2Pre-trained=true2024.0583.3
HPSv22026.0583.3
MPSBackbone=CLIP2026.0283.27
UnifiedRewardBackbone=LLaVA-OV-7B2026.0283.1
UnifiedReward-ThinkBackbone=LLaVA-OV-7B2026.0282.7
HPSv2Backbone=CLIP2026.0282.58
DRMWeights=Pre-trained, Epoch=1, Size=5122026.0582.2
DiNa-LRMBackbone=SD3.5-M-2B, Multi-noise inference scaling=false2026.0282.13
HPSv3 - 2B2026.0580.8
DRMWeights=Pre-trained, Epoch=1, Size=2562026.0580.1
PickScorePre-trained=true2024.0579.8
PickScore2026.0579.8
PickScoreBackbone=CLIP2026.0279.44
HPS2026.0577.6
Aesthetic Score Predictor2026.0576.8
ImageReward2026.0574
ImageRewardBackbone=CLIP2026.0273.95
HPSPre-trained=true2024.0573.1
Aesthetic ScorePre-trained=true2024.0572.6
LRM-SD1.5Backbone=SD-1.52026.0272.39
CLIP scorePre-trained=true2024.0571.2
LRM-SDXLBackbone=SDXL2026.0271.19
ImageRewardPre-trained=true2024.0570.6
DRMWeights=Random, Epoch=3, Size=2562026.0570.1
DRMWeights=Random, Epoch=2, Size=2562026.0568.5
CLIP ViT-H/142026.0565.1
DRMWeights=Random, Epoch=1, Size=2562026.0565