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

Multi-image reasoning on Mantis (test)

72.81Accuracy

DelimScaling

Updated 1mo ago

Evaluation Results

Method	Links
DelimScaling 2026.02		72.81
InternVL3 2026.02		71.89
DelimScaling 2026.02		70.05
DelimScaling 2026.02		69.12
DelimScaling 2026.02		69.12
Qwen2.5-VL 2026.02		68.66
Qwen2.5-VL 2026.02		68.2
InternVL3 2026.02		67.28
LLaVA-NeXT-Interleave-14B 2024.09		66.4
MM1.5-30B 2024.09		64.6
DelimScaling 2026.02		64.06
DelimScaling 2026.02		63.13
LLaVA-NeXT-Interleave-7B 2024.09		62.7
GPT-4V 2024.09		62.7
LLaVA-OV 2026.02		62.21
Qwen2.5-VL 2026.02		59.91
MM1.5-7B 2024.09		57.6
Mantis-Idefics2-8B 2024.09		57.1
BLIP-3-4B 2024.09		56.7
MM1.5-3B 2024.09		54.8
MM1.5-3B-MoE 2024.09		54.4
DelimScaling 2026.02		54.38
MM1-30B 2024.09		52.5
InternVL3 2026.02		52.07
MM1-7B 2024.09		51.6
MM1.5-1B-MoE 2024.09		51.2
MM1.5-1B 2024.09		50.7
DelimScaling 2026.02		49.77
Idefics2-8B 2024.09		48.9
Phi-3-Vision-4B 2024.09		47.9
InternVL3 2026.02		47
LLaVA-NeXT-Interleave-0.5B 2024.09		45.6
MM1-3B 2024.09		45.2
MM1-1B 2024.09		41.5
DelimScaling 2026.02		41.01
LLaVA-OV 2026.02		40.09
LLaVAOneVision-0.5B 2024.09		39.6
Emu2-Chat-37B 2024.09		37.8
LLaVA-v1.5-7B 2024.09		31.3