List your company
ResearchFormerly WizWand

Visual Question Answering on TextVQA v1.0 (val)

85.5AccuracyQwen2-VL-72B
-3.191219.834442.8665.8856Apr 18, 2019Jun 12, 2020Aug 8, 2021Oct 4, 2022Nov 29, 2023Jan 24, 2025Mar 22, 2026
Updated 7mo ago

Evaluation results

MethodLinks
Qwen2-VL-72B2024.1285.5
Human2019.0485.01
InternVL2-Llama3-76B2024.1284.4
Qwen2-VL-7B2024.1284.3
InternVL2.5-78B2024.1283.4
Molmo-72B2024.1283.1
InternVL2-40B2024.1283
InternVL2.5-38B2024.1282.7
InternVL2.5-26B2024.1282.4
InternVL2-26B2024.1282.3
NVLM-D-72B2024.1282.1
Molmo-7B-D2024.1281.7
InternVL-Chat-V1.52024.1280.6
LLaVA-OneVision-72B2024.1280.5
MiniCPM-V2.62024.1280.1
Qwen2-VL-2B2024.1279.7
InternVL2.5-8B2024.1279.1
Gemini-1.5-Pro2024.1278.8
GPT-4V2024.1278
InternVL2-8B2024.1277.4
GPT-4o-202405132024.1277.4
InternVL2.5-4B2024.1276.8
Cambrian-34B2024.1276.7
Aquila-VL-2B2024.1276.4
InternVL2-4B2024.1274.4
InternVL2.5-2B2024.1274.3
Claude-3.5-Sonnet2024.1274.1
VILA-1.5-40B2024.1273.6
InternVL2-2B2024.1273.4
InternVL2.5-1B2024.1272
Phi-3.5-Vision-4B2024.1272
InternVL2-1B2024.1270.5
LA+OCR UBVocab=LA2019.0467.56
Claude-3-Opus2024.1267.5
MobileVLM V2 7BLLM=Vicuna-7B, #Samples=3.6M2026.0362.3
LaTr+-LargeOCR System=Amazon-OCR, Pre-Training Data=IDL, Extra Finetune=ST-VQA2021.1261.05
ShareGPT4VLLM=Vicuna-7B, #Samples=1.9M2026.0360.4
LaTr-LargeOCR System=Amazon-OCR, Pre-Training Data=IDL, Extra Finetune=None2021.1259.76
LaTr+-BaseOCR System=Amazon-OCR, Pre-Training Data=IDL, Extra Finetune=ST-VQA2021.1259.53
LLaVA-1.5LLM=Vicuna-7B, #Samples=1.2M2026.0358.2
LaTr-BaseOCR System=Amazon-OCR, Pre-Training Data=IDL, Extra Finetune=None2021.1258.03
TAPOCR System=Microsoft-OCR, Pre-Training Data=TextVQA, ST-VQA, TextCaps, OCR-CC, Extra Finetune=ST-VQA2021.1254.71
Cosine KD w/ Beta-KD (Instance)LLM=MobileLLaMA 1.4B, #Samples=3.6M2026.0354.2
Align-KD w/ Beta-KD (Instance)LLM=MobileLLaMA 1.4B, #Samples=3.6M2026.0353.9
Cosine KD w/ Beta-KD (Task)LLM=MobileLLaMA 1.4B, #Samples=3.6M2026.0353.9
Align-KD w/ Beta-KD (Task)LLM=MobileLLaMA 1.4B, #Samples=3.6M2026.0352.9
Align-KDLLM=MobileLLaMA 1.4B, #Samples=3.6M2026.0352.8
LaTr-BaseOCR System=Amazon-OCR, Pre-Training Data=None, Extra Finetune=None2021.1252.29
Cosine KDLLM=MobileLLaMA 1.4B, #Samples=3.6M2026.0352.2
LOGOSOCR System=Microsoft-OCR, Pre-Training Data=None, Extra Finetune=ST-VQA2021.1251.53
MobileVLM -V2 1.7BLLM=MobileLLaMA 1.4B, #Samples=3.6M2026.0351.2
TAPOCR System=Microsoft-OCR, Pre-Training Data=TextVQA, ST-VQA, Extra Finetune=ST-VQA2021.1250.57
TAPOCR System=Microsoft-OCR, Pre-Training Data=TextVQA, Extra Finetune=None2021.1249.91
LA UBVocab=LA2019.0448.46
LaTr-BaseOCR System=Rosetta-en, Pre-Training Data=IDL, Extra Finetune=None2021.1248.38
MoE-LLaVA-2BLLM=Qwen-1.5-1.8B, #Samples=2.2M2026.0348
M4COCR System=Amazon-OCR, Pre-Training Data=None, Extra Finetune=None2021.1247.84
MoE-LLaVA-1.6BLLM=StableLM-1.6B, #Samples=2.2M2026.0347.8
MobileVLM 3BLLM=MobileLLaMA 2.7B, #Samples=3.6M2026.0347.5
SA-M4COCR System=Google-OCR, Pre-Training Data=None, Extra Finetune=ST-VQA2021.1245.4
LLAVADILLM=MobileLLaMA 1.4B, #Samples=3.6M2026.0345.3
M4COCR System=Microsoft-OCR, Pre-Training Data=None, Extra Finetune=ST-VQA2021.1245.22
TAPOCR System=Rosetta-en, Pre-Training Data=TextVQA, Extra Finetune=None2021.1244.06
LaTr-BaseOCR System=Rosetta-en, Pre-Training Data=None, Extra Finetune=None2021.1244.06
MobileVLM 1.7BLLM=MobileLLaMA 1.4B, #Samples=3.6M2026.0341.5
LaAP-NetOCR System=Rosetta-en, Pre-Training Data=None, Extra Finetune=None2021.1240.68
CRNOCR System=Rosetta-en, Pre-Training Data=None, Extra Finetune=None2021.1240.39
SMAOCR System=Rosetta-en, Pre-Training Data=None, Extra Finetune=None2021.1240.05
M4COCR System=Rosetta-en, Pre-Training Data=None, Extra Finetune=None2021.1239.4
OCR UBDescription=Upper bound from OCR tokens2019.0437.12
LoRRA (Pythia backbone)Vocab=SA2019.0426.56
LoRRA (Pythia backbone)Vocab=LA2019.0426.23
Pythia (I+Q) + O + CVocab=n/a, Features=Image + Question + OCR + Copy Mechanism2019.0420.06
LoRRA (BAN backbone)Vocab=SA2019.0418.41
Pythia (I+Q) + OVocab=LA, Features=Image + Question + OCR features2019.0418.35
Pythia (I+Q)Vocab=LA, Features=Image + Question2019.0413.04
BAN (I+Q)Vocab=LA2019.0412.3
OCR Max2019.049.76
QVocab=LA, Features=Question only2019.048.09
Random OCR2019.047.72
IVocab=LA, Features=Image only2019.046.29
Majority Ans2019.044.48
Wt. Rand 1002019.040.27
Rand 1002019.040.22