List your company

Visual Grounding on RefCOCO (TestB)

92.6AccuracyVanilla
77.842481.673785.50589.3363Mar 30, 2022Dec 1, 2022Aug 5, 2023Apr 7, 2024Dec 10, 2024Aug 13, 2025Apr 17, 2026

Evaluation results

MethodLinks
VanillaBackbone Model=DeepSeek-VL2 Small-16B, Average Tokens=100%2026.0292.6———————
InternVL2.5Number of Parameters=78B2025.0292.5———————
Florence-2Backbone=DaViT-B2026.0192———————
Florence-2-L#PT=126M, #FT=127K2025.0392———————
IVC-PruneBackbone Model=DeepSeek-VL2 Small-16B, Average Tokens=52%2026.0291.8———————
UNINEXT#PT=600K, #FT=127K2025.0391.5———————
Florence-2-B#PT=126M, #FT=127K2025.0391.5———————
UNINEXTModel Category=Specialist2024.0391.46———————
Qwen3-VL-8B-Thinking + EGMBackbone=Qwen3-VL-8B, Enhancement=EGM2026.0191.2———————
Qwen3-VL-4B-Thinking + EGMBackbone=Qwen3-VL-4B, Enhancement=EGM2026.0190.9———————
Youtu-VLModel Size=4B2026.0190.8———————
Qwen3-VL-235B-A22B-ThinkingBackbone=Qwen3-VL-235B-A22B, Variant=Thinking2026.0190.6———————
W-FastVBackbone Model=DeepSeek-VL2 Small-16B, Average Tokens=54%2026.0290.4———————
InternVL-3-78BBackbone=InternVL-3-78B2026.0190.3———————
Qwen3-VL-235B-A22B-InstructBackbone=Qwen3-VL-235B-A22B, Variant=Instruct2026.0190.2———————
InternVL-3-8B + EGMBackbone=InternVL-3-8B, Enhancement=EGM2026.0190.1———————
Qwen2.5-VLNumber of Parameters=72B2025.0289.7———————
FastVBackbone Model=DeepSeek-VL2 Small-16B, Average Tokens=54%2026.0289.5———————
Qwen3-VL-2B-Thinking + EGMBackbone=Qwen3-VL-2B, Enhancement=EGM2026.0189.4———————
InfMLLM-7BLLM=Vicuna-7B2023.1189.3———————
ONE-PEACE2023.0589.26———————
UFOBackbone=InternVL2.5-8B2026.0189.2———————
PDropBackbone Model=DeepSeek-VL2 Small-16B, Average Tokens=57%2026.0289.1———————
CogVLM-Grounding-17BModel Size=17B, Zero-shot=false2024.0288.99———————
Ferretv2#PT=1.1M, #FT=127K2025.0388.7———————
OFA2023.0588.44———————
Qwen2.5-VL-7B AIFAIF=true2026.0488.4———————
Grounding-DINO2023.0588.24———————
G-DINO-LModel Category=Specialist2024.0388.24———————
Grounding DINO2025.0288.2———————
Grounding DINOBackbone=Swin-L2026.0188.2———————
InternVL-3.5Model Size=4B2026.0188.2———————
Grounding-DINO-L2026.0488.2———————
InternVL-3-8BBackbone=InternVL-3-8B2026.0188———————
Qwen3-VL-8B-InstructBackbone=Qwen3-VL-8B, Variant=Instruct2026.0187.8———————
Qwen3-VL-4B-InstructBackbone=Qwen3-VL-4B, Variant=Instruct2026.0187.6———————
InternVL-3-2B + EGMBackbone=InternVL-3-2B, Enhancement=EGM2026.0187.4———————
Phi3V-dist#PT=0, #FT=127K2025.0387.3———————
FIBER2023.0587.26———————
InternVL-3-1B + EGMBackbone=InternVL-3-1B, Enhancement=EGM2026.0187———————
Qwen3-VL-8B-ThinkingBackbone=Qwen3-VL-8B, Variant=Thinking2026.0186.9———————
Qwen3-VLModel Size=4B2026.0186.7———————
InternVL-3-2BBackbone=InternVL-3-2B2026.0186.4———————
GriffonBackbone=LLama2-13B2026.0186.1———————
VanillaBackbone Model=InternVL 2.5 8B, Average Tokens=100%2026.0286———————
Phi3V-sft#PT=0, #FT=127K2025.0386———————
Phi3V-vocab#PT=0, #FT=127K2025.0386———————
InternVL-2.5-8B2026.0485.9———————
IVC-PruneBackbone Model=InternVL 2.5 8B, Average Tokens=50%2026.0285.7———————
Qwen3-VL-4B-ThinkingBackbone=Qwen3-VL-4B, Variant=Thinking2026.0185.6———————
Qwen2.5-VLNumber of Parameters=7B2025.0285.4———————
Qwen2.5-VL-7B2026.0485.4———————
QWen-VL-7BLLM=Qwen-7B2023.1185.34———————
Qwen-VL-7BModel Size=7B, Zero-shot=false2024.0285.34———————
MiniGPT-v2-7BLLM=LLaMA2-7B2023.1185.33———————
VanillaBackbone=QWEN-2.5-VL 7B, Average Tokens=100%2026.0285.16———————
VanillaBackbone Model=Qwen2.5-VL 7B, Average Tokens=100%2026.0284.7———————
Qwen3-VL-2B-InstructBackbone=Qwen3-VL-2B, Variant=Instruct2026.0184.7———————
IVC-PruneBackbone Model=Qwen2.5-VL 7B, Average Tokens=50%2026.0284.5———————
Ferret#PT=1.1M, #FT=127K2025.0384.4———————
NuwaBackbone=QWEN-2.5-VL 7B, Average Tokens=75%2026.0284.37———————
VisionLLM v2Backbone=Swin-T2026.0184.3———————
Qwen2.5-VLNumber of Parameters=3B2025.0284———————
SeqTRVisual Encoder=DN53, Params (M)=7.90, Pre-train images=174K2022.0383.59———————
Lumen-7BModel Category=Generalist2024.0383.55———————
Localization Heads2026.0483.3———————
UNICORN2023.0583.06———————
VistaLLM-7BModel Size=7B, Zero-shot=false2024.0283———————
NuwaBackbone=QWEN-2.5-VL 7B, Average Tokens=50%2026.0282.85———————
MDETRBackbone=ENB32026.0182.7———————
Ferret-7BModel Size=7B, Zero-shot=false2024.0282.45———————
CoLLaVO-7BModel Size=7B, Zero-shot=true2024.0282.39———————
QRNet2024.0182.34———————
VScanBackbone Model=Qwen2.5-VL 7B, Average Tokens=50%2026.0282.2———————
Qwen3-VL-2B-ThinkingBackbone=Qwen3-VL-2B, Variant=Thinking2026.0182———————
Shikra-13BLLM=Vicuna-13B2023.1181.81———————
Shikra2026.0281.8———————
UVLTrackModel Version=B2024.0181.73———————
InternVL-3-1BBackbone=InternVL-3-1B2026.0181.7———————
MDETRVisual Encoder=RN101, Params (M)=17.36, Pre-train images=200K2022.0381.41———————
MDETR2023.0581.41———————
SeqTR2024.0181.24———————
Fully-FTArchitecture=MDETR2023.0881.22———————
Fully-FTParams.=185.2, Mem.=39.62026.0481.22———————
Fully-FTParams. (M)=185.2, Mem. (G)=39.62026.0481.22———————
RefTRVisual Encoder=RN101, Params (M)=17.86, Pre-train images=100K2022.0381.16———————
APE (D)Backbone=ViT-L2023.1280.968.4——————
InternVL2-8B2026.0480.7———————
VLTVG2024.0180.49———————
Shikra-7BLLM=Vicuna-7B2023.1180.24———————
Shikra-7BModel Category=Generalist2024.0380.24———————
Shikra-7BModel Size=7B, Zero-shot=false2024.0280.24———————
Shikra-7B2026.0480.2———————
W-FastVBackbone Model=Qwen2.5-VL 7B, Average Tokens=54%2026.0279.8———————
Partially⇑Architecture=MDETR2023.0879.78———————
Ours†Params. (M)=0.8, Mem. (G)=7.32026.0479.2———————
Ours‡Params. (M)=0.8, Mem. (G)=7.42026.0479.19———————
Ours♠Params.=0.8, Mem.=7.1, Time (ms)=52.292026.0478.97———————
Ours♡Params.=0.7, Mem.=7.1, Time (ms)=52.632026.0478.69———————
SHERLParams.=0.7, Mem.=6.8, Time (ms)=79.332026.0478.41———————