List your company
ResearchFormerly WizWand

Open-ended Generation on LLaVA-Bench

84.3GPT-4 ScoreLACING
50.08458.96767.8576.733Nov 21, 2024
Updated 5mo ago

Evaluation results

MethodLinks
LACINGModel Size=13B2024.1184.3
LLaVA1.5-13B-BPOModel Size=13B, Baseline Category=reinforcement learning-based2024.1174.4
VCDModel Size=13B, Baseline Category=training-free2024.1173.6
VDD-NoneModel Size=13B, Baseline Category=training-free2024.1173
LLaVA-1.5Model Size=13B, Baseline Category=LVLMs2024.1172.5
ICDModel Size=13B, Baseline Category=training-free2024.1172.5
LACINGModel Size=7B2024.1172.2
LLaVA1.5-7B-BPOModel Size=7B, Baseline Category=reinforcement learning-based2024.1171.6
LLaVAModel Size=13B, Baseline Category=LVLMs2024.1170.8
MuffinModel Size=13B, Baseline Category=LVLMs2024.1168.8
HA-DPOModel Size=7B, Baseline Category=reinforcement learning-based2024.1167.2
VDD-NoneModel Size=7B, Baseline Category=training-free2024.1166
VCDModel Size=7B, Baseline Category=training-free2024.1165.3
ICDModel Size=7B, Baseline Category=training-free2024.1164.7
LLaVA-1.5Model Size=7B, Baseline Category=LVLMs2024.1164.4
POVIDModel Size=7B, Baseline Category=reinforcement learning-based2024.1162.2
LLaVA-RLHFModel Size=13B, Baseline Category=reinforcement learning-based2024.1161.5
Less-is-moreModel Size=7B, Baseline Category=training-free2024.1160.9
OPERAModel Size=7B, Baseline Category=training-free2024.1160.3
QWEN-VLModel Size=10B, Baseline Category=LVLMs2024.1152.1
RLHF-VModel Size=13B, Baseline Category=reinforcement learning-based2024.1151.4