List your company
ResearchFormerly WizWand

Visual Question Answering on TextVQA (TextVQA Score)

90TextVQA ScoreSPOT-E
-2.97621.16245.369.438Dec 11, 2024Mar 15, 2025Jun 18, 2025Sep 21, 2025Dec 24, 2025Mar 29, 2026Jul 2, 2026
Updated 4mo ago

Evaluation results

MethodLinks
SPOT-EBase Model=Qwen3-VL-32B2026.0690
SPOT-EBase Model=Qwen2.5-VL-32B2026.0689.5
Qwen3-VL-32BBase Model=Qwen3-VL-32B2026.0689
Qwen2.5-VL-32BBase Model=Qwen2.5-VL-32B2026.0688.3
SPOT-EBase Model=LLaVA-OV-72B2026.0688
LLaVA-OV-72BBase Model=LLaVA-OV-72B2026.0686.4
SPOT-EBase Model=InternVL2.5-26B2026.0686.1
InternVL2.5-26BBase Model=InternVL2.5-26B2026.0684.6
Full PrecisionModel=Qwen2.5-VL-7B-Instruct, Weight bits=16 bit, Memory=15.45 GB, Evaluation protocol=zero-shot2026.0783.88
Qwen3-VL-8BRetained Tokens=1024 Tokens (100%)2026.0783.8
Full PrecisionModel=Qwen2.5-VL-72B-Instruct, Weight bits=16 bit, Memory=136.74 GB, Evaluation protocol=zero-shot2026.0783.26
GPTQModel=Qwen2.5-VL-7B-Instruct, Weight bits=3 bit, Memory=2.66 GB, Evaluation protocol=zero-shot2026.0782.25
GPTQModel=Qwen2.5-VL-72B-Instruct, Weight bits=3 bit, Memory=28.61 GB, Evaluation protocol=zero-shot2026.0782.03
InternVL3-8BBackbone=InternVL3-8B, Retained Tokens=1280, Reduction Ratio=0%2026.0681.5
SAB-LVLMModel=Qwen2.5-VL-72B-Instruct, Weight bits=1.07 bit, Memory=23.23 GB, Evaluation protocol=zero-shot2026.0780.76
ARB-LLMModel=Qwen2.5-VL-72B-Instruct, Weight bits=1.19 bit, Memory=23.23 GB, Evaluation protocol=zero-shot2026.0780.23
Full PrecisionModel=Qwen2.5-VL-32B-Instruct, Weight bits=16 bit, Memory=62.31 GB, Evaluation protocol=zero-shot2026.0778.85
BiLLMModel=Qwen2.5-VL-72B-Instruct, Weight bits=1.08 bit, Memory=24.23 GB, Evaluation protocol=zero-shot2026.0777.28
SAB-LVLMModel=Qwen2.5-VL-32B-Instruct, Weight bits=1.07 bit, Memory=10.28 GB, Evaluation protocol=zero-shot2026.0775.45
EADPRetained Tokens=512 Tokens (↓ 50.0%)2026.0775.2
DivPruneRetained Tokens=512 Tokens (↓ 50.0%)2026.0775
ARB-LLMModel=Qwen2.5-VL-32B-Instruct, Weight bits=1.16 bit, Memory=10.28 GB, Evaluation protocol=zero-shot2026.0774.96
CDPrunerRetained Tokens=512 Tokens (↓ 50.0%)2026.0774.5
GPTQModel=Qwen2.5-VL-32B-Instruct, Weight bits=3 bit, Memory=12.71 GB, Evaluation protocol=zero-shot2026.0774.45
PB-LLMModel=Qwen2.5-VL-72B-Instruct, Weight bits=1.70 bit, Memory=22.99 GB, Evaluation protocol=zero-shot2026.0774.34
SAB-LVLMModel=Qwen2.5-VL-7B-Instruct, Weight bits=1.07 bit, Memory=2.14 GB, Evaluation protocol=zero-shot2026.0774
EADPRetained Tokens=256 Tokens (↓ 75.0%)2026.0771.4
ERABackbone=InternVL3-8B, Retained Tokens=256, Reduction Ratio=80.0%2026.0671.2
ARB-LLMModel=Qwen2.5-VL-7B-Instruct, Weight bits=1.07 bit, Memory=2.14 GB, Evaluation protocol=zero-shot2026.0770.7
DivPruneRetained Tokens=256 Tokens (↓ 75.0%)2026.0769.5
CDPrunerRetained Tokens=256 Tokens (↓ 75.0%)2026.0768.5
VisionZipBackbone=InternVL3-8B, Retained Tokens=256, Reduction Ratio=80.0%2026.0667.2
DivPruneBackbone=InternVL3-8B, Retained Tokens=256, Reduction Ratio=80.0%2026.0667.1
EADPRetained Tokens=128 Tokens (↓ 87.5%)2026.0766.1
PB-LLMModel=Qwen2.5-VL-32B-Instruct, Weight bits=1.70 bit, Memory=10.24 GB, Evaluation protocol=zero-shot2026.0763.62
DivPruneRetained Tokens=128 Tokens (↓ 87.5%)2026.0762.6
DARTBackbone=InternVL3-8B, Retained Tokens=256, Reduction Ratio=80.0%2026.0662
CleanModel=LLaVA-1.5, Vision Encoder=CLIP-3362024.1261.3
ERABackbone=InternVL3-8B, Retained Tokens=128, Reduction Ratio=90.0%2026.0660.8
CDPrunerRetained Tokens=128 Tokens (↓ 87.5%)2026.0760.6
CleanModel=Qwen2.5-VL, Vision Encoder=Qwen-ViT2024.1259.9
Full-DataSampling Ratio=100% (665K), Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0557.9
Both-EmbModel=LLaVA-1.5, Vision Encoder=CLIP-3362024.1256.2
CoIDOSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0556
OFASampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0556
BiLLMModel=Qwen2.5-VL-32B-Instruct, Weight bits=1.08 bit, Memory=10.73 GB, Evaluation protocol=zero-shot2026.0755.92
Img-EmbModel=LLaVA-1.5, Vision Encoder=CLIP-3362024.1255.4
PreSelSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0555.2
XMASSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0555.2
ICONSSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0555.2
COINCIDESampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0554.8
DivPruneBackbone=InternVL3-8B, Retained Tokens=128, Reduction Ratio=90.0%2026.0654.7
RandomSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0554.3
Text-EmbModel=LLaVA-1.5, Vision Encoder=CLIP-3362024.1254.1
CLIP-ScoreSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0553.4
BiLLMModel=Qwen2.5-VL-7B-Instruct, Weight bits=1.08 bit, Memory=2.24 GB, Evaluation protocol=zero-shot2026.0753.38
TypiClustSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0553.3
IFDSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0551.8
Self-FilterSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0551.4
DARTBackbone=InternVL3-8B, Retained Tokens=128, Reduction Ratio=90.0%2026.0651.3
EL2NSampling Ratio=15%, Backbone=LLaVA-v1.5-7B, Training Dataset=LLaVA-665K2026.0550.2
CleanModel=LLaVA, Vision Encoder=CLIP-3362024.1250.2
VisionZipBackbone=InternVL3-8B, Retained Tokens=128, Reduction Ratio=90.0%2026.0648.3
Img-EmbModel=LLaVA, Vision Encoder=CLIP-3362024.1248
Img-EmbModel=Qwen2.5-VL, Vision Encoder=Qwen-ViT2024.1247.7
Both-EmbModel=LLaVA, Vision Encoder=CLIP-3362024.1247.4
Text-EmbModel=LLaVA, Vision Encoder=CLIP-3362024.1244.8
VEV-UAPModel=Qwen2.5-VL, Vision Encoder=Qwen-ViT2024.1243.8
CleanModel=InstructBlip, Vision Encoder=EVA-CLIP2024.1232.5
Img-EmbModel=InstructBlip, Vision Encoder=EVA-CLIP2024.1230.1
Both-EmbModel=InstructBlip, Vision Encoder=EVA-CLIP2024.1229.6
PB-LLMModel=Qwen2.5-VL-7B-Instruct, Weight bits=1.17 bit, Memory=2.15 GB, Evaluation protocol=zero-shot2026.0728.67
Text-EmbModel=InstructBlip, Vision Encoder=EVA-CLIP2024.1226.6
VEV-UAPModel=InstructBlip, Vision Encoder=EVA-CLIP2024.1226.3
VEV-UAPModel=LLaVA, Vision Encoder=CLIP-3362024.122.9
VEV-UAPModel=LLaVA-1.5, Vision Encoder=CLIP-3362024.120.6