List your company
ResearchFormerly WizWand

Personalized Image Aesthetic Assessment on PARA

0.62Spearman's RhoPIAA-ICI (w/o User Attr)
0.180080.294290.40850.52271Apr 13, 2026Apr 14, 2026Apr 15, 2026Apr 16, 2026Apr 17, 2026Apr 18, 2026Apr 19, 2026
Updated 6mo ago

Evaluation results

MethodLinks
PIAA-ICI (w/o User Attr)User Attributes=false2026.040.6239.2———
PIAA-ICI (w/ User Attr)User Attributes=true2026.040.61942.4———
Linear-HiddenBackbone=Qwen3-VL 4B, LT layers=152026.040.61136.2———
Linear-HiddenSupport=100-shot, Backbone=Qwen3-VL, Model Size=4B2026.040.6110.362———
Linear-HiddenSupport=100-shot, Backbone=Qwen3-VL, Model Size=2B2026.040.6040.363———
Linear-Hidden (GIAA)Support=100-shot, Backbone=Qwen3-VL, Model Size=4B2026.040.6030.057———
Linear-Hidden (Reduce)Support=100-shot, Backbone=Qwen3-VL, Model Size=4B2026.040.5970.382———
Linear-Hidden (GIAA)Support=100-shot, Backbone=Qwen3-VL, Model Size=2B2026.040.5960.041———
Linear-Hidden (GIAA)Support=100-shot, Backbone=Qwen3-VL, Model Size=8B2026.040.5960.043———
Linear-HiddenSupport=100-shot, Backbone=Gemma 3, Model Size=12B2026.040.5940.329———
Linear-Hidden (GIAA)Support=100-shot, Backbone=Gemma 3, Model Size=12B2026.040.5940.036———
Linear-Hidden (Reduce)Support=100-shot, Backbone=Gemma 3, Model Size=12B2026.040.5930.373———
Linear-Hidden (Reduce)Support=100-shot, Backbone=Gemma 3, Model Size=4B2026.040.5920.365———
Linear-HiddenSupport=100-shot, Backbone=Qwen3-VL, Model Size=8B2026.040.5910.341———
Linear-HiddenSupport=100-shot, Backbone=Gemma 3, Model Size=4B2026.040.5910.346———
Linear-Hidden (Reduce)Support=100-shot, Backbone=Qwen3-VL, Model Size=2B2026.040.5850.367———
Linear-Hidden (GIAA)Support=100-shot, Backbone=Gemma 3, Model Size=4B2026.040.5840.014———
LoRASupport=100-shot, Backbone=Qwen3-VL, Model Size=4B2026.040.5781.751———
Raw TextBackbone=Qwen3-VL, Model Size=4B2026.040.571.277———
Adjust-BiasSupport=100-shot, Backbone=Qwen3-VL, Model Size=4B2026.040.570.672———
LoRASupport=100-shot, Backbone=Qwen3-VL, Model Size=8B2026.040.5680.978———
Linear-Hidden (Reduce)Support=100-shot, Backbone=Qwen3-VL, Model Size=8B2026.040.5580.322———
P-MLLMSize=8B, Zero-shot=true2026.040.557—0.0190.6080.013
Raw TextBackbone=Qwen3-VL, Model Size=8B2026.040.5280.729———
Adjust-BiasSupport=100-shot, Backbone=Qwen3-VL, Model Size=8B2026.040.5280.441———
LoRASupport=100-shot, Backbone=Gemma 3, Model Size=12B2026.040.5240.525———
P-MLLM-SSize=8B, Zero-shot=true2026.040.518—0.0140.5570.017
Raw TextBackbone=Qwen3-VL, Model Size=2B2026.040.5040.571———
Adjust-BiasSupport=100-shot, Backbone=Qwen3-VL, Model Size=2B2026.040.5040.31———
AesExpertSize=7B, Zero-shot=true2026.040.498—0.0190.5140.019
GPT-4o-miniZero-shot=true2026.040.494—0.0260.5090.025
Raw TextBackbone=Gemma 3, Model Size=12B2026.040.4931.879———
Adjust-BiasSupport=100-shot, Backbone=Gemma 3, Model Size=12B2026.040.4931.562———
LoRASupport=100-shot, Backbone=Gemma 3, Model Size=4B2026.040.4890.893———
LoRASupport=100-shot, Backbone=Qwen3-VL, Model Size=2B2026.040.4871.97———
Raw TextBackbone=Gemma 3, Model Size=4B2026.040.4621.107———
Adjust-BiasSupport=100-shot, Backbone=Gemma 3, Model Size=4B2026.040.4620.321———
Few-shotSupport=10-shot, Backbone=Gemma 3, Model Size=12B2026.040.4070.185———
Q-InstructSize=7B, Zero-shot=true2026.040.39—0.0180.4990.02
Few-shotSupport=10-shot, Backbone=Qwen3-VL, Model Size=8B2026.040.3720.547———
Qwen2.5-VLSize=7B, Zero-shot=true2026.040.357—0.020.3690.019
Few-shotSupport=10-shot, Backbone=Qwen3-VL, Model Size=2B2026.040.3191.85———
Few-shotSupport=10-shot, Backbone=Gemma 3, Model Size=4B2026.040.2410.537———
Few-shotSupport=10-shot, Backbone=Qwen3-VL, Model Size=4B2026.040.1971.576———