List your company
ResearchFormerly WizWand

Visual Question Answering on VizWizQA

78.9AccuracyPragma-VL
32.72444.71256.768.688Aug 9, 2024Nov 11, 2024Feb 14, 2025May 20, 2025Aug 22, 2025Nov 25, 2025Feb 28, 2026
Updated 7mo ago

Evaluation results

MethodLinks
Pragma-VLBackbone=Qwen2.5-VL-7B2026.0278.9
DPOBackbone=Qwen2.5-VL-7B2026.0273.81
Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.0272.53
SPA-VLBackbone=Qwen2.5-VL-7B2026.0271.65
DPOBackbone=Llava-1.5-7B2026.0269.2
SFTBackbone=Llava-1.5-7B2026.0269.03
Llava-1.5-7BBackbone=Llava-1.5-7B2026.0268.93
Beavertails-V_helpBackbone=Qwen2.5-VL-7B2026.0268.85
SFTBackbone=Qwen2.5-VL-7B2026.0268.64
MM-RLHFBackbone=Qwen2.5-VL-7B2026.0268.07
Pragma-VLBackbone=Llava-1.5-7B2026.0267.78
Beavertails-V_harmBackbone=Llava-1.5-7B2026.0266.69
Yi-VL-6BFinetuning Dataset=Origin2025.0466.2
MM-RLHFBackbone=Llava-1.5-7B2026.0266.14
LLaVA-NeXT-Mistral-7BFinetuning Dataset=Origin2025.0464.98
Beavertails-V_harmBackbone=Qwen2.5-VL-7B2026.0264.26
mPLUG-Owl3# Param=8B2024.0863.5
SPA-VLBackbone=Llava-1.5-7B2026.0262.59
Beavertails-V_helpBackbone=Llava-1.5-7B2026.0260.07
Yi-VL-6BFinetuning Dataset=VLGuard2025.0459.32
LLaVA-v1.5-13BFinetuning Dataset=Ours2025.0458.77
VILA-1.5# Param=8B2024.0858.7
LLAVA-Next# Param=8B2024.0857.6
LLaVA-v1.5-13BFinetuning Dataset=Origin2025.0457.34
Flamingo# Param=80B2024.0857.2
LLaVA-v1.5-13BFinetuning Dataset=VLGuard2025.0456.77
LLaVA-v1.5-7BFinetuning Dataset=VLGuard2025.0456.36
LLaVA-v1.5-7BFinetuning Dataset=Ours2025.0455.92
LLaVA-NeXT-Mistral-7BFinetuning Dataset=Ours2025.0455.5
LLaVA-v1.5-7BFinetuning Dataset=Origin2025.0455.1
mPLUG-Owl2# Param=8B2024.0854.5
LLaVA-NeXT-Mistral-7BFinetuning Dataset=VLGuard2025.0453.15
LLAVA-1.5# Param=8B2024.0850
Yi-VL-6BFinetuning Dataset=Ours2025.0447.91
EVLM-Chat# Param=32B2024.0847.3
Qwen-VL-Chat# Param=9B2024.0838.9
InstructBLIP# Param=7B2024.0834.5