List your company
ResearchFormerly WizWand

Visual Perception on AI2D

84AccuracyINTERNVL3-8B + PGT
51.55259.97668.476.824Mar 18, 2026Mar 28, 2026Apr 8, 2026Apr 19, 2026Apr 30, 2026May 11, 2026May 22, 2026
Updated 5mo ago

Evaluation results

MethodLinks
INTERNVL3-8B + PGTBackbone=INTERNVL3-8B, Data=PGT-augmented2026.0584
INTERNVL3-8BBackbone=INTERNVL3-8B2026.0583.8
THINKLITE-VLMethod=THINKLITE-VL2026.0583.4
QWEN2.5-VL-7B + PGTBackbone=QWEN2.5-VL-7B, Data=PGT-augmented2026.0583.2
Qwen2-VLSize=7B2026.0383
IMAGE JIGSAWMethod=IMAGE JIGSAW2026.0582.9
QWEN2.5-VL-7BBackbone=QWEN2.5-VL-7B2026.0582.7
QWEN2.5-VL-7B + SPECIALIZED MIXBackbone=QWEN2.5-VL-7B, Data=Specialized Mix2026.0582.7
Ovis1.5-LLaMA3Size=8B2026.0382.5
Qwen2.5-VLSize=7B2026.0382.4
MiniCPM-V-2.6Size=7B2026.0382.1
Insight-V++Size=7B, Base Model=Qwen2.5-VL, Self-Evolving=true2026.0381.7
Qwen2.5-VL + Multi-Agent (RL)Size=7B, Multi-Agent (RL)=true2026.0381.4
InternLM-XComposer2Size=7B2026.0381.2
POINTSSize=7B2026.0381.2
Our Base Model + Multi-AgentSize=7B, Multi-Agent=true2026.0380.1
Insight-VSize=7B, Base Model=Our Base Model, Iterative DPO=true2026.0379.8
Our Base ModelSize=7B2026.0379.7
QWEN2.5-VL-3B + SPECIALIZED MIXBackbone=QWEN2.5-VL-3B, Data=Specialized Mix2026.0579.6
QWEN2.5-VL-3BBackbone=QWEN2.5-VL-3B2026.0579.2
DPA-Qwen3-32BTarget LLM=Qwen3-32B2026.0579
MiniCPM-LLaMA3-V 2.5Size=8B2026.0378.4
QWEN2.5-VL-3B + PGTBackbone=QWEN2.5-VL-3B, Data=PGT-augmented2026.0578.1
Insight-V-LLaVASize=8B, Base Model=LLaVA-NeXT-LLaMA3, Iterative DPO=true2026.0377.3
Idefics3-LLaMA3Size=8B2026.0376.5
LLaVA-NeXT-LLaMA3 + Multi-AgentSize=8B, Multi-Agent=true2026.0375.7
Qwen2-VL-2BTarget LLM=Public Models2026.0574.7
Cambrian-1Size=8B2026.0374.6
Cambrian-1-8BTarget LLM=Public Models2026.0574.6
SPATIAL-LADDER-3BMethod=SPATIAL-LADDER-3B2026.0573.5
LLaVA-NeXT-Qwen3-32BTarget LLM=Qwen3-32B2026.0573
Idefics2-8BTarget LLM=Public Models2026.0572.5
LLaVA-NeXT-LLaMA3Size=8B2026.0371.5
LLAVA-NEXT-LLAMA3-8BBackbone=LLAVA-NEXT-LLAMA3-8B2026.0571.5
LLAVA-NEXT-LLAMA3-8B + PGTBackbone=LLAVA-NEXT-LLAMA3-8B, Data=PGT-augmented2026.0571.3
VIGORL-3BMethod=VIGORL-3B2026.0571.2
DPA-Qwen3-4BTarget LLM=Qwen3-4B2026.0570
Bunny-LLaMA3Size=8B2026.0369.4
LLaVA-NeXT-Qwen3-4BTarget LLM=Qwen3-4B2026.0567.4
LLaVA-NeXT-7BTarget LLM=Public Models2026.0566.6
DeepSeek-VLSize=7B2026.0365.3
LLAVA-NEXT-7BBackbone=LLAVA-NEXT-7B2026.0564.5
LLAVA-NEXT-7B + PGTBackbone=LLAVA-NEXT-7B, Data=PGT-augmented2026.0564.5
DPA-LLaMA-3.2-3BTarget LLM=LLaMA-3.2-3B2026.0559.5
LLaVA-NeXT-LLaMA-3.2-3BTarget LLM=LLaMA-3.2-3B2026.0559
VILA-1.5Size=8B2026.0358.8
LLaVA-1.5-7BTarget LLM=Public Models2026.0552.8