Share your thoughts, 1 month free Claude Pro on usSee more

Text-based Visual Question Answering on TextVQA (Accuracy)

88.5Accuracy

Q-Mask

Updated 3d ago

Evaluation Results

Method
Q-Mask 2026.03	88.5	-	-
Qwen2.5-VL-32B + VPPO 2026.02	86.2	-	-
ToolTree 2026.03	85.43	-	-
DualPD 2026.01	85.34	-	-
Qwen2.5-VL-32B + AT-RL (Ours) 2026.02	85.3	-	-
Qwen2.5VL 2026.02	84.9	-	-
Qwen2.5-VL 2025.10	84.9	-	-
Qwen2.5-VL 2026.03	84.9	-	-
DualPD 2026.01	84.84	-	-
Qwen2.5-VL-7B 2026.04	84.8	-	-
Qwen2.5-VL-32B Instruct 2026.02	84.6	-	-
LLaVA-OV 2026.02	84.5	-	-
Qwen2.5-VL-72B Instruct 2026.02	84.3	-	-
Manzano 2026.02	84.3	-	-
Qwen2-VL 2025.10	84.3	-	-
InternVL3.5 2026.03	84.3	-	-
DoLA 2026.01	84.19	-	-
DoLA 2026.01	83.99	-	-
Qwen-2.5-VL-7B 2026.01	83.72	-	-
Qwen2.5-VL 2026.03	83.5	-	-
Qwen-2-VL-7B 2026.01	83.34	-	-
Molmo 2026.03	83.1	-	-
Dense 2026.03	82.9	-	-
MBQ 2026.03	82.9	-	-
MBQ 2026.03	82.9	-	-
InternVL3.5 2026.03	82.7	-	-
SQ 2026.03	82.6	-	-
MASQuant 2026.03	82.6	-	-
Q-Mask 2026.03	82.6	-	-
FP16 2024.12	82.5	-	78.1
MBQ 2024.12	82.5	-	77.6
MASQuant 2026.03	82.5	-	-
RTN 2026.03	82.5	-	-
AWQ 2024.12	82.4	-	77.5
ToolTree 2026.03	82.26	-	-
Full Model 2026.02	82.24	-	-
AWQ 2026.03	82.2	-	-
FP16 2024.12	82.1	-	74.6
RTN 2026.03	82.1	-	-
Qwen3-VL 2026.03	82.1	-	-
Qwen2VL-7B 2025.04	81.7	-	-
GPTQ 2024.12	81.6	-	76.6
Kelix 2026.02	81.4	-	-
MBQ 2024.12	81.1	-	73.8
Wanda 2026.02	81.08	-	-
AWQ 2024.12	81	-	73.5
Qwen3-VL + DISCO (10K) 2026.02	80.83	-	-
POP 2026.02	80.73	-	-
RTN 2024.12	80.6	-	73.3
VQ-RAE 2026.02	80.6	-	-
Qwen3-VL 2026.03	80.6	-	-
MBQ 2024.12	80.5	-	75.8
SpaRE-7B 2025.04	80.5	-	-
OpenAI GPT-4o 2026.02	80.5	-	-
Qwen3-VL 2026.02	80.34	-	-
InternVL3 2025.10	80.2	-	-
GPTQ 2024.12	80.1	-	72.3
DualPD 2026.01	80.1	-	-
MBQ 2024.12	80	-	72.7
Bagel 2026.02	80	-	-
Qwen3-VL + Textual (10K) 2026.02	79.91	-	-
RTN 2024.12	79.7	-	75
Qwen2.5VL 2026.02	79.7	-	-
Qwen2-VL 2025.10	79.7	-	-
RTN 2024.12	79.6	-	72.7
GLM-4.1V 2026.03	79.6	-	-
DeSAP 2026.04	79.5	-	-
FP16 2024.12	79.3	-	74.3
Qwen2.5-VL 3B (reported) 2025.12	79.3	-	-
Qwen2.5-VL 2025.10	79.3	-	-
TokenVL-8B 2026.03	79.3	-	-
Qwen2.5-VL 2026.03	79.3	-	-
Qwen3-VL 2026.03	79.3	-	-
SpaRE-2B 2025.04	79.2	-	-
Qwen2.5-VL 3B (reproduced) 2025.12	79.2	-	-
InternVL2.5 2025.10	79.1	-	-
HoloV 2026.04	78.9	-	-
MBQ 2024.12	78.5	-	73.6
InternVL3.5 2026.03	78.5	-	-
Oryx-1.5 2024.09	78.3	-	-
LLaVA-1.6-13B 2026.01	78.2	-	-
DoLA 2026.01	78.2	-	-
InternVL3.5 2026.03	78.2	-	-
Gemini 2.0 Flash 2026.02	78.1	-	-
Blip-3o 2026.02	78	-	-
CD 2026.01	77.96	-	-
GPTQ 2024.12	77.9	-	72.3
InternVL3.5 2026.02	77.9	-	-
DualPD 2026.01	77.9	-	-
Dense 2026.03	77.9	-	-
InternVL3.5 2026.03	77.9	-	-
FastV 2026.04	77.9	-	-
LongVILA-7B (S3) 2024.08	77.8	-	-
Oryx 2024.09	77.8	-	-
Qwen2.5-VL 2026.02	77.8	-	-
MBQ 2026.03	77.8	-	-
Qwen-2.5-VL-7B 2025.08	77.72	-	-
Vanilla 2026.02	77.7	-	-
MASQuant 2026.03	77.7	-	-
InternVL2-8B 2025.04	77.6	-	-

Showing 100 of 817 rows

...