List your company

Fine-grained Visual Reasoning on V* OOD

90Overall ScoreDeepEyes
62.85669.90376.9583.997Jul 8, 2026

Evaluation results

MethodLinks
DeepEyes2026.079092.186.8
TikArtBase Model=Qwen3-VL-8B2026.0789.591.386.8
SIEVEBase Model=Qwen3-VL-8B2026.078888.786.8
TreeVGR2026.0787.489.584.2
MIRROR2026.0783.8——
BUS-GRPOBase Model=Qwen3-VL-8B2026.0783.885.281.6
BUS-SFTBase Model=Qwen3-VL-8B2026.0782.783.581.6
Thyme2026.0782.283.580.3
DeepEyesV22026.0781.881.780.3
V-Reflection2026.0781.783.578.9
PixelReasoner2026.0780.683.576.3
Gemini-2.5-flash2026.0780.184.473.7
Gemini-2.5-pro2026.0779.184.471.1
Qwen3-VL-8B2026.0777.580.273.7
GPT-52026.0776.474.879
Qwen3-VL-4B2026.0774.978.369.7
Qwen2.5-VL-7B2026.0774.377.469.7
InternVL3-8B2026.0772.37371.1
GPT-4o2026.0770.771.369.7
GPT-5-nano2026.0763.958.372.4