List your company

Referring Expression Counting on REC 8K (val)

4.23MAECAD-GD
3.127210.571118.01525.4589May 15, 2026May 20, 2026May 25, 2026May 31, 2026Jun 5, 2026Jun 10, 2026Jun 16, 2026

Evaluation results

MethodLinks
CAD-GDBackbone=Swin-B2026.064.2313.14
Qwen2.5VL-7B-Instruct + REC-RLTraining Dataset=Subset-2K2026.055.0515.87
DCountBackbone=Swin-T2026.065.2415.47
MambaCount2026.065.2816.26
Qwen2.5VL-7B-Instruct + FGRPRTraining Dataset=Subset-2K2026.055.2916.2
CAD-GDBackbone=Swin-T2026.065.4315.01
Qwen2.5VL-3B-Instruct + REC-RLTraining Dataset=Subset-2K2026.055.4816.27
Qwen2.5VL-3B-Instruct + FGRPRTraining Dataset=Subset-2K2026.055.6517.41
Ground-D RECTraining Dataset=all data2026.056.818.13
GroundingREC2026.056.818.13
GroundingRECBackbone=Swin-T2026.066.818.13
Qwen2.5VL-7B-Instruct + SFTTraining Dataset=all data2026.058.2720.64
Qwen2.5VL-3B-Instruct + SFTTraining Dataset=all data2026.058.6322.66
Qwen2.5VL-7B-InstructTraining Dataset=None2026.059.0222.6
Ground-D Swin-TTraining Dataset=all data2026.059.0321.98
GDino2026.059.0321.98
GroundingDINOBackbone=Swin-T2026.069.0321.98
Qwen2.5VL-3B-InstructTraining Dataset=None2026.059.2422.54
VisHarnesstraining=GRPO2026.059.3123.16
CountGD2026.059.5122.91
VisHarness-SFTtraining=SFT2026.059.8928.16
Ground-D Swin-TTraining Dataset=None2026.0511.7728.6
CounTX2026.0511.8827.04
CountTX2026.0611.8827.04
ZSC (Swin-T)Backbone=Swin-T2026.0512.9626.74
ZSCBackbone=Swin-T2026.0612.9626.74
Mean2026.0514.2827.75
Mean2026.0614.2827.75
ZSC (ResNet-50)Backbone=ResNet-502026.0514.8431.3
ZSCBackbone=ResNet-502026.0614.8431.3
TFOC2026.0516.0831.61
Qwen3.5-9BModel size=9B2026.0527.4103.5
Qwen3-VL-8BModel size=8B2026.0531.8142.5