List your company

Fine-grained Visual Question Answering on HRBench 4K

71.13Overall AccuracyFOCUS
33.034842.924952.81562.7051Jun 26, 2025

Evaluation results

MethodLinks
FOCUSBackbone=LLaVA-OV-7B, Steps=8, Overrun=true2025.0671.1386.7555.513.41
FOCUSBackbone=LLaVA-OV-7B, Steps=4, Overrun=true2025.0670.7585.7555.759.32
FOCUSBackbone=LLaVA-OV-7B, Steps=6, Overrun=true2025.0670.6385.7555.511.33
FOCUSBackbone=LLaVA-OV-7B, Steps=2, Overrun=false2025.0670.388456.756.07
FOCUSBackbone=LLaVA-OV-7B, Steps=2, Overrun=true2025.0670.2586.5547.41
FOCUSBackbone=LLaVA-OV-7B, Steps=3, Overrun=true2025.067085.7554.258.34
FOCUSBackbone=LLaVA-OV-7B, Steps=1, Overrun=true2025.0669.8885.75546.55
ZoomEyeBackbone=LLaVA-OV-7B, Depth=5, Number of crops=42025.0669.3884.554.2529.19
FOCUSBackbone=LLaVA-OV-7B, Steps=1, Overrun=false2025.066982.2555.753.84
ZoomEyeBackbone=LLaVA-OV-7B, Depth=2, Number of crops=42025.0666.1380.255220.42
ZoomEyeBackbone=LLaVA-OV-7B, Depth=1, Number of crops=42025.0665.7580.7550.7511.43
ZoomEyeBackbone=LLaVA-OV-7B, Depth=1, Number of crops=22025.0661.387052.759.46
FOCUSBackbone=LLaVA-1.5-7B, Steps=8, Overrun=true2025.0651.756439.514.49
FOCUSBackbone=LLaVA-1.5-7B, Steps=6, Overrun=true2025.0650.6362.253912.31
ZoomEyeBackbone=LLaVA-1.5-7B, Depth=5, Number of crops=42025.065063.2536.7549.38
FOCUSBackbone=LLaVA-1.5-7B, Steps=4, Overrun=true2025.0649.2559.7538.7510.14
ZoomEyeBackbone=LLaVA-1.5-7B, Depth=2, Number of crops=42025.0647.7557.2538.2535.6
FOCUSBackbone=LLaVA-1.5-7B, Steps=3, Overrun=true2025.0647.1356.537.759.09
FOCUSBackbone=LLaVA-1.5-7B, Steps=2, Overrun=false2025.0646.556375.41
FOCUSBackbone=LLaVA-1.5-7B, Steps=2, Overrun=true2025.0645.8855.75365.95
ViCropBackbone=LLaVA-1.5-7B, Mode=grad-att-high2025.0644.2553.7534.7515.8
ViCropBackbone=LLaVA-1.5-7B, Mode=rel-att-high2025.0642.551.533.530.59
DC2Backbone=LLaVA-v.1.5-7B2025.0642.3——48.55
FOCUSBackbone=LLaVA-1.5-7B, Steps=1, Overrun=true2025.0642.2551.5334.99
FOCUSBackbone=LLaVA-1.5-7B, Steps=1, Overrun=false2025.0641.1349.7532.53.14
ZoomEyeBackbone=LLaVA-1.5-7B, Depth=1, Number of crops=42025.0639.1344.7533.517.46
ZoomEyeBackbone=LLaVA-1.5-7B, Depth=1, Number of crops=22025.0636.2539.2533.2511.54
SEAL2025.0634.5——18.05