List your company
ResearchFormerly WizWand

GUI Visual Question Answering and Grounding on MMBench-GUI L2

80.4AccuracyGROUNDNEXT-7B (SFT)
32.0444.59557.1569.705Nov 10, 2025
Updated 4mo ago

Evaluation results

MethodLinks
GROUNDNEXT-7B (SFT)Model Scale=7B, Training Paradigm=SFT2025.1180.4
GROUNDNEXT-3B (SFT)Model Scale=3B, Training Paradigm=SFT2025.1175.5
GUI-Actor-7BModel Scale=7B, Training Paradigm=SFT2025.1170.9
JEDI-7BModel Scale=7B, Training Paradigm=SFT2025.1170.4
GUI-Actor-3BModel Scale=3B, Training Paradigm=SFT2025.1169.8
Qwen2.5-VL-7B (Agent mode)Model Scale=7B, Evaluation Mode=Agent mode, Training Paradigm=SFT2025.1167.7
JEDI-3BModel Scale=3B, Training Paradigm=SFT2025.1166.5
UGround-V1-7BModel Scale=7B, Training Paradigm=SFT2025.1165.7
Qwen2.5-VL-3BModel Scale=3B, Training Paradigm=SFT2025.1160.8
Qwen2.5-VL-3B (Agent mode)Model Scale=3B, Evaluation Mode=Agent mode, Training Paradigm=SFT2025.1160.8
PhiGround-4B-7CModel Scale=4B, Training Paradigm=SFT2025.1160.3
Aguvis-7BModel Scale=7B, Training Paradigm=SFT2025.1145.7
OS-Atlas-7BModel Scale=7B, Training Paradigm=SFT2025.1141.4
Qwen2.5-VL-7BModel Scale=7B, Training Paradigm=SFT2025.1133.9