Share your thoughts, 1 month free Claude Pro on usSee more

Multimodal Reasoning on SPD-Faith Bench Hard 1.0

10.9Contradiction Rate

MiniCPM-V-2.6

Updated 1mo ago

Evaluation Results

Method	Links
MiniCPM-V-2.6 2026.02		10.9
Qwen2.5-VL-72B 2026.02		11.4
InternVL2.5-38B 2026.02		11.8
Qwen2.5-VL-7B 2026.02		12.2
InternVL2.5-8B 2026.02		12.5
Qwen3-VL-32B 2026.02		12.9
Qwen3-VL-235B-A22B 2026.02		17
GLM-4.5V 2026.02		19.5
GPT-4o 2026.02		22.8
Gemini-2.5-Pro 2026.02		27.2
DeepSeek-VL2 2026.02		32.5
Claude-4.5-Haiku 2026.02		37.5