Referring Expression Counting on REC 8K (test)
4.34MAECAD-GD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CAD-GDBackbone=Swin-B2026.06 | 4.34 | 12.93 | |
| Qwen2.5VL-7B-Instruct + REC-RLTraining Dataset=Subset-2K2026.05 | 5.06 | 13.92 | |
| CAD-GDBackbone=Swin-T2026.06 | 5.29 | 17.08 | |
| MambaCount2026.06 | 5.42 | 17.18 | |
| DCountBackbone=Swin-T2026.06 | 5.45 | 18.61 | |
| Qwen2.5VL-3B-Instruct + REC-RLTraining Dataset=Subset-2K2026.05 | 5.48 | 16.51 | |
| Qwen2.5VL-7B-Instruct + FGRPRTraining Dataset=Subset-2K2026.05 | 5.67 | 23.21 | |
| Qwen2.5VL-3B-Instruct + FGRPRTraining Dataset=Subset-2K2026.05 | 5.86 | 19 | |
| Ground-D RECTraining Dataset=all data2026.05 | 6.5 | 19.79 | |
| GroundingREC2026.05 | 6.5 | 19.79 | |
| GroundingRECBackbone=Swin-T2026.06 | 6.5 | 19.79 | |
| GrRECBackbone=Swin-T, FT=true2026.06 | 6.5 | 19.79 | |
| ABACUS-3BBackbone=UniLIP-3B, FT=false2026.06 | 7.67 | 15.84 | |
| Ground-D Swin-TTraining Dataset=all data2026.05 | 8.88 | 21.95 | |
| GDino2026.05 | 8.88 | 21.95 | |
| GroundingDINOBackbone=Swin-T2026.06 | 8.88 | 21.95 | |
| GDinoBackbone=Swin-T, FT=true2026.06 | 8.88 | 21.95 | |
| Qwen2.5VL-3B-Instruct + SFTTraining Dataset=all data2026.05 | 8.9 | 25.04 | |
| Qwen2.5VL-7B-Instruct + SFTTraining Dataset=all data2026.05 | 9.05 | 26.23 | |
| Qwen2.5VL-7B-InstructTraining Dataset=None2026.05 | 9.1 | 28.18 | |
| Qwen2.5VL-3B-InstructTraining Dataset=None2026.05 | 9.67 | 27.98 | |
| VisHarnesstraining=GRPO2026.05 | 9.86 | 24.7 | |
| VisHarness-SFTtraining=SFT2026.05 | 9.92 | 27.1 | |
| CountGD2026.05 | 11.33 | 30.87 | |
| Ground-D Swin-TTraining Dataset=None2026.05 | 11.71 | 26.97 | |
| CounTX2026.05 | 11.84 | 25.62 | |
| CountTX2026.06 | 11.84 | 25.62 | |
| CounTXBackbone=ViT-B/16, FT=true2026.06 | 11.84 | 25.62 | |
| ZSC (Swin-T)Backbone=Swin-T2026.05 | 13 | 29.07 | |
| ZSCBackbone=Swin-T2026.06 | 13 | 29.07 | |
| ZSCBackbone=Swin-T, FT=true2026.06 | 13 | 29.07 | |
| Mean2026.05 | 13.75 | 25.91 | |
| Mean2026.06 | 13.75 | 25.91 | |
| UniLIP-3BFT=false2026.06 | 13.75 | 25.91 | |
| ZSC (ResNet-50)Backbone=ResNet-502026.05 | 14.93 | 29.72 | |
| ZSCBackbone=ResNet-502026.06 | 14.93 | 29.72 | |
| TFOC2026.05 | 17.27 | 32.68 | |
| TFOCBackbone=ViT-B, FT=false2026.06 | 17.27 | 32.68 | |
| Qwen3.5-9BModel size=9B2026.05 | 28.9 | 97.8 | |
| Qwen3-VL-8BModel size=8B2026.05 | 35.7 | 152.3 |