List your company

Android UI Control on AndroidControl-CL Task Order 1 v1 (test)

77.83Shopping AccuracySFT
65.152468.443771.73575.0263Mar 3, 2026

Evaluation results

MethodLinks
SFTBackbone=LLava-onsevision-0.5b, Training Strategy=Joint-Training2026.0377.8384.4683.2478.4184.7669.7379.5879.6928.63—
RIF-RFTBackbone=LLava-onsevision-0.5b2026.0376.5374.1975.9973.5581.465.9175.3474.422.17-0.88
CGLBackbone=LLava-onsevision-0.5b2026.0375.5581.3280.7376.8883.4168.978.1377.8424.77-0.52
GRPOBackbone=LLava-onsevision-0.5b2026.0375.4478.8678.0176.4181.5767.9377.876.5723.85-0.93
SFT+KLBackbone=LLava-onsevision-0.5b2026.0374.1675.2376.8975.4982.0666.0179.2975.5922.39-1.78
SFT+ReplayBackbone=LLava-onsevision-0.5b2026.0370.8876.4377.2374.1978.2164.3780.6374.5620.07-4.69
SFTBackbone=LLava-onsevision-0.5b2026.0365.6474.7374.667277.3964.1280.0872.6614.61-6.81