List your company
ResearchFormerly WizWand

Multi-image reasoning on BLINK (val)

52.6AccuracyLLaVA-NeXT-Interleave-7B
35.54439.97244.448.828Sep 30, 2024
Updated 8mo ago

Evaluation results

MethodLinks
LLaVA-NeXT-Interleave-7BModel Scale=7B2024.0952.6
LLaVAOneVision-0.5BModel Scale=1B2024.0952.1
LLaVA-NeXT-Interleave-14BModel Scale=>14B2024.0952.1
GPT-4VModel Scale=>14B2024.0951.1
MM1.5-30BModel Scale=>14B2024.0950.2
MM1.5-3B-MoEModel Scale=3B2024.0949.8
BLIP-3-4BModel Scale=3B2024.0949.7
Mantis-Idefics2-8BModel Scale=7B2024.0949.1
MM1.5-7BModel Scale=7B2024.0948.2
MM1.5-3BModel Scale=3B2024.0946.8
MM1.5-1BModel Scale=1B2024.0946.3
Idefics2-8BModel Scale=7B2024.0945.2
Phi-3-Vision-4BModel Scale=3B2024.0944.2
MM1.5-1B-MoEModel Scale=1B2024.0943.7
MM1-30BModel Scale=>14B2024.0943.5
MM1-3BModel Scale=3B2024.0941.5
MM1-1BModel Scale=1B2024.0940.3
MM1-7BModel Scale=7B2024.0940
LLaVA-NeXT-Interleave-0.5BModel Scale=1B2024.0939.2
LLaVA-v1.5-7BModel Scale=7B2024.0937.1
Emu2-Chat-37BModel Scale=>14B2024.0936.2