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

Multimodal Evaluation on LLaVA-Bench-Wild (LLaVA-W)

97.3Overall Score

MoE-LLaVA

Updated 18d ago

Evaluation Results

Method	Links
MoE-LLaVA 2024.02		97.3
MoE-LLaVA 2024.02		94.1
VILA-13B + ShareGPT4V 2023.12		78.4
TinyLLaVA-share-Sig-Phi 2024.02		75.8
VILA-13B 2023.12		73
LLaVA-1.5 2023.12		70.7
VILA-7B 2023.12		69.7
Full-Finetune 2026.05		69.5
Self-Filter 2026.05		69.1
MAGIC 2026.05		69
Perplexity 2026.05		68.7
SemDedup 2026.05		68.7
Random 2026.05		68.6
CLIP-Score 2026.05		68.1
COINCIDE 2026.05		67.4
TinyLLaVA-share-C-Phi 2024.02		67.1
D2-Pruning 2026.05		66.5
PADE 2026.02		65.92
EL2N 2026.05		65.8
V-ITI 2026.02		65.44
TrustLLaVA 2026.07		65.3
Self-Sup 2026.05		64.9
Vanilla 2026.02		64.8
OPERA 2026.02		64.31
LLaVA-1.5 2026.07		64.2
HyperLLaVA 2024.03		64
HyperLLaVA w/o Ev 2024.03		63.7
LLaVA-1.5 2024.02		63.4
LLaVA-1.5 2024.03		63.4
LLaVA-1.5 2023.12		63.4
LLaVA-1.5 2026.07		63.4
Full-Finetune 2026.05		63.3
VCD 2026.02		63.21
MAGIC 2026.05		63.2
HyperLLaVA w/o EL 2024.03		62.6
INTER 2026.02		61.7
LLaVA-SP 2026.07		61.7
InstructBLIP 2024.02		60.9
InstructBLIP 2023.12		60.9
Random 2026.05		58.7
ICD 2026.02		56.9
TrustLLaVA (MLP) 2026.07		51.5
BLIP-2 2023.12		38.1