List your company

Long-context Visual Question Answering on Visual HayStack Short Window (test)

98.2Accuracy (1 Image)SFT(LoRA)
77.077682.561388.04593.5287Dec 25, 2025

Evaluation results

MethodLinks
SFT(LoRA)Model Parameters=3B2025.1298.291.8887.4868.8931.5
SFT(LoRA)Model Parameters=7B2025.1297.7892.9285.884.7335.92
LAidModel Parameters=3B2025.1296.8383.3474.2963.2720.1
LAidModel Parameters=7B2025.1292.8383.2680.4674.0924.1
YaRNModel Parameters=3B2025.1286.2772.8657.7955.742.8
BaseModel Parameters=3B2025.1285.9165.762.0952.16—
BaseModel Parameters=32B2025.1283.7979.1174.7173.34—
BaseModel Parameters=7B2025.1280.2268.4562.1957.21—
YaRNModel Parameters=7B2025.1280.0363.7862.0955.96-2.5
SelfExtendModel Parameters=7B2025.1278.5362.5858.6953.01-5.9
SelfExtendModel Parameters=3B2025.1277.8964.6954.1147.95-7.9