List your company

Fine-grained Visual Reasoning on MME-RW-Lite ID

57Overall ScoreTikArt
41.71245.68149.6553.619Jul 8, 2026

Evaluation results

MethodLinks
TikArtBase Model=Qwen3-VL-8B2026.075760.553.5
SIEVEBase Model=Qwen3-VL-8B2026.0756.458.453.2
GPT-52026.0756.256.955.5
TreeVGR2026.0754.958.249.7
BUS-GRPOBase Model=Qwen3-VL-8B2026.0754.457.649.5
Thyme2026.0754.458.248.7
BUS-SFTBase Model=Qwen3-VL-8B2026.0754.258.148.1
V-Reflection2026.0753.958.545
Gemini-2.5-pro2026.0753.552.150.7
DeepEyes2026.0753.255.446.8
DeepEyesV22026.0752.454.848
Gemini-2.5-flash2026.0752.149.747.3
MIRROR2026.0751.5——
PixelReasoner2026.0749.75345.6
Qwen3-VL-8B2026.0748.65440.4
InternVL3-8B2026.0747.951.142.9
Qwen3-VL-4B2026.0747.151.839.7
GPT-4o2026.0745.749.853.9
GPT-5-nano2026.0742.846.149.4
Qwen2.5-VL-7B2026.0742.346.535.9