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

Visual Perception on AI2D

83Accuracy

Qwen2-VL

Updated 29d ago

Evaluation Results

Method	Links
Qwen2-VL 2026.03		83
Ovis1.5-LLaMA3 2026.03		82.5
Qwen2.5-VL 2026.03		82.4
MiniCPM-V-2.6 2026.03		82.1
Insight-V++ 2026.03		81.7
Qwen2.5-VL + Multi-Agent (RL) 2026.03		81.4
InternLM-XComposer2 2026.03		81.2
POINTS 2026.03		81.2
Our Base Model + Multi-Agent 2026.03		80.1
Insight-V 2026.03		79.8
Our Base Model 2026.03		79.7
MiniCPM-LLaMA3-V 2.5 2026.03		78.4
Insight-V-LLaVA 2026.03		77.3
Idefics3-LLaMA3 2026.03		76.5
LLaVA-NeXT-LLaMA3 + Multi-Agent 2026.03		75.7
Cambrian-1 2026.03		74.6
LLaVA-NeXT-LLaMA3 2026.03		71.5
Bunny-LLaMA3 2026.03		69.4
DeepSeek-VL 2026.03		65.3
VILA-1.5 2026.03		58.8