List your company
ResearchFormerly WizWand

Vision Understanding on CVBench-2D

77.76AccuracyJigsaw + CARE
42.857651.918860.9870.0412Dec 1, 2024Feb 2, 2025Apr 6, 2025Jun 9, 2025Aug 11, 2025Oct 13, 2025Dec 16, 2025
Updated 8mo ago

Evaluation results

MethodLinks
Jigsaw + CAREModel Size=7B2025.1277.76
Mix + CL + CAREModel Size=7B2025.1276.96
Jigsaw + CL + CAREModel Size=7B2025.1275.52
Jigsaw + CLModel Size=7B2025.1275.37
Rotation + CL + CAREModel Size=7B2025.1275.08
GRPO-CAREModel Size=7B2025.1274.91
Visual JigsawModel Size=7B2025.1274.46
VisualSphinxModel Size=7B2025.1273.98
Qwen 2.5 VLModel Size=7B2025.1273.62
Vision-ZeroModel Size=7B2025.1273.53
PatchFit + CL + CAREModel Size=7B2025.1273.37
JigsawModel Size=7B2025.1272.13
ViCritModel Size=7B2025.1270.84
LLaVA-1.5-13BImage (prefill) Token=576, TFLOPs=19.62024.1262.3
Dynamic-LLaVA-13B^ITFree=false, Image (prefill) Token=115, TFLOPs=4.72024.1260.7
LLaVA-FastV (13B)Image (prefill) Token=144, TFLOPs=62024.1258.9
LLaVA-1.5-7BImage (prefill) Token=576, TFLOPs=10.12024.1258.5
Dynamic-LLaVA-7B^ITFree=false, Image (prefill) Token=115, TFLOPs=2.52024.1258.3
LLaVA-1.5-13b+H2Ok-10,-0.5Image (prefill) Token=576, TFLOPs=19.62024.1250.8
LLaVA-FastV† (7B)Free=false, Image (prefill) Token=144, TFLOPs=3.22024.1249.8
LLaVA-1.5-7B+H2Ok-10, -0.5Image (prefill) Token=576, TFLOPs=10.12024.1249.7
LLaVA-HiRED (7B)Free=true, Image (prefill) Token=115, TFLOPs=22024.1244.2