List your company

Multimodal Reasoning on VLMsAreBlind

52.32Score (%)Qwen2.5-VL
47.286448.593249.951.2068Jun 1, 2026

Evaluation results

MethodLinks
Qwen2.5-VLScale=7B2026.0652.32
LLaVA-OVScale=1.5-8B2026.0651.07
MOSS-Video-PreviewSFT Mode=real-time2026.0650.21
MOSS-Video-PreviewSFT Mode=offline2026.0647.48