List your company

Visual Question Answering on CLEVR (val)

100Overall AccuracyXNM-GT
49.35262.50175.6588.799Aug 1, 2018Nov 7, 2019Feb 12, 2021May 21, 2022Aug 27, 2023Dec 2, 2024Mar 11, 2026

Evaluation results

MethodLinks
XNM-GTProgram=GT, #Modules=12, #Param.=0.22M2018.12100100100100100100————
XNM-GTProgram=supervised, #Modules=12, #Param.=0.22M2018.1299.999.999.899.999.899.9————
NS-VQAProgram=supervised, #Modules=122018.1299.899.799.999.999.899.8————
NSVQAPre-training=Visual Pre-training2022.0199.7—————————
S-MACPre-training=Visual Pre-training2022.0198.9—————————
XNM-DetProgram=GT, #Modules=12, #Param.=0.55M2018.1297.996.298.898.198.797.8————
XNM-DetProgram=supervised, #Modules=12, #Param.=0.55M2018.1297.89698.698.198.797.8————
FiLMPre-training=Visual Pre-training2022.0197.7—————————
Stack-NMNs**Expert layout=true2018.0896.6—————————
Object RN**2018.0894.593.694.793.395.294.4————
VQATransformerPre-training=None2022.0193.6—————————
Stack-NMNs**Expert layout=false2018.0893—————————
HumanPre-training=Visual Pre-training2022.0192.6—————————
VQATransformer + LexSymPre-training=None2022.0192—————————
VQATransformer + SubstitutePre-training=None2022.0190.8—————————
SNs (mean)Backbone=Qwen3-VL-4b-Instruct2026.0388.3—————88.387.99089
SNs (maj. voting)Backbone=Qwen3-VL-4b-Instruct2026.0388.3—————88.385.392.588.7
Qwen3-VL-4b-Instruct2026.0383.7——————76.796.885.6
SNs (mean)Backbone=LLaVA-v1.5-7b2026.0351.5—————535175.861
SNs (maj. voting)Backbone=LLaVA-v1.5-7b2026.0351.4—————5350.977.561.4
LLaVA-v1.5-7b2026.0351.3——————50.976.661.1