List your company
ResearchFormerly WizWand

Visual Question Answering on VQAT

0.736AccuracyDragonfly
0.213920.349460.4850.62054Feb 8, 2024Jul 2, 2024Nov 25, 2024Apr 19, 2025Sep 12, 2025Feb 4, 2026Jun 30, 2026
Updated 4mo ago

Evaluation results

MethodLinks
DragonflyBackbone=Llama3-8B, #Data=2.9M2024.060.736
VILA2-8BLLM Parameters=8B, Vision Tower Parameters=400M, Tokens per image=196, Pre-training data size=51M2024.070.734
Idefics2-8BLLM Parameters=8B, Vision Tower Parameters=400M, Tokens per image=320, Pre-training data size=>600M2024.070.73
MM1-7B-ChatLLM Parameters=7B, Vision Tower Parameters=300M, Tokens per image=720, Pre-training data size=>2B2024.070.728
CogVLM-13B-ChatBackbone=Vicuna-7B, #Data=>1.5B2024.060.704
LLaVA-UHDBackbone=Vicuna-13B, #Data=1.2M2024.060.677
LLaVA-NeXTBackbone=Vicuna-13B, #Data=1.2M2024.060.671
VILABackbone=Llama2-13B, #Data=51M2024.060.666
Qwen-VLObserved in training=true2024.020.638
LLaVA-1.5Backbone=Vicuna-13B, #Data=1.2M2024.060.613
LLaVA-1.5 + QA-ViTObserved in training=true2024.020.591
InternVL-13BBackbone=Vicuna-13B, #Data=6B2024.060.587
LLaVA-1.5-7BVision Token=5762026.060.582
DARTVision Token=1922026.060.574
LLaVA-1.5Observed in training=true2024.020.574
FitPruneVision Token=1922026.060.574
MS-ResamplerVision Token=1442026.060.572
MustDropVision Token=1922026.060.565
PDropVision Token=1922026.060.561
SparseVLMVision Token=1922026.060.561
VanillaVision Token=576, Backbone=Qwen2.5-3B, Vision Encoder=clip-vit-large-patch14-3362026.060.547
LLaVA-PruMergeVision Token=1922026.060.543
mPLUG-DocOwlObserved in training=true2024.020.526
FastVVision Token=1922026.060.525
MS-ResamplerVision Token=144, Backbone=Qwen2.5-3B, Vision Encoder=clip-vit-large-patch14-3362026.060.524
TokenPackerVision Token=144, Backbone=Qwen2.5-3B, Vision Encoder=clip-vit-large-patch14-3362026.060.517
Pixel-ShuffleVision Token=144, Backbone=Qwen2.5-3B, Vision Encoder=clip-vit-large-patch14-3362026.060.513
LDP-v2Vision Token=144, Backbone=Qwen2.5-3B, Vision Encoder=clip-vit-large-patch14-3362026.060.511
InstructBLIPBackbone=Vicuna-13B, #Data=130M2024.060.507
C-AbstractorVision Token=144, Backbone=Qwen2.5-3B, Vision Encoder=clip-vit-large-patch14-3362026.060.48
HiREDVision Token=1922026.060.474
InstructBLIPOCR tokens=true2024.020.466
InstructBLIP2024.020.309
IDEFICS-80B2024.020.309
IDEFICS-9B2024.020.259
OpenFlamingo-9B2024.020.242
BLIP22024.020.234