List your company

GUI Navigation on Online-Mind2Web (Hard)

53.52Success RateClaude 4 Sonnet CU
7.697619.593831.4943.3862Feb 14, 2026

Evaluation results

MethodLinks
Claude 4 Sonnet CUModel Type=Proprietary2026.0253.52
Ovis2.5S-GRPOTraining Method=GRPO (Step-wise)2026.0251.35
Ovis2.5S-RLOOTraining Method=RLOO (Step-wise)2026.0251.35
Claude 3.7 Sonnet CUModel Type=Proprietary2026.0246.48
Ovis2.5S-RF++Training Method=REINFORCE++ (Step-wise)2026.0244.59
Ovis2.5SFTTraining Method=SFT2026.0239.19
UI-TARSModel Scale=7B, Version=1.52026.0217.57
Qwen3-VLModel Scale=32B2026.0214.86
Qwen3-VLModel Scale=8B2026.029.46