List your company

Hallucination Evaluation on MME hallucination (test)

180Existence ScoreVASparse
41.336877.3359113.335149.3341Mar 27, 2024May 14, 2024Jul 1, 2024Aug 19, 2024Oct 6, 2024Nov 23, 2024Jan 11, 2025

Evaluation results

MethodLinks
VASparseBackbone=LLaVA-1.52025.01180132.67121.33160—
VCDBackbone=LLaVA-1.52025.01175.33130.33115155—
VASparseBackbone=mPLUG-Owl22025.01175130110.67155—
VCDBackbone=mPLUG-Owl22025.01171.33125107.33150—
DoLaBackbone=LLaVA-1.52025.01170120106.67150.67—
HALCBackbone=LLaVA-1.5, Post-hoc=true2025.01167.67121.33106.67150.67—
GreedyBackbone=mPLUG-Owl22025.01167120105145—
DoLaBackbone=mPLUG-Owl22025.01167125110147.67—
OPERABackbone=mPLUG-Owl22025.01167122.33100145—
HALCBackbone=mPLUG-Owl2, Post-hoc=true2025.01167120.33108.67145—
GreedyBackbone=LLaVA-1.52025.01165.67120110.67148.33—
OPERABackbone=LLaVA-1.52025.01165115.67104145—
VASparseBackbone=MiniGPT-42025.01147.3398.6778.67133—
VCDBackbone=MiniGPT-42025.0114295.3371.33129—
OPERABackbone=MiniGPT-42025.01140.6792.3373125—
HALCBackbone=MiniGPT-4, Post-hoc=true2025.0114092.6771.33122.67—
GreedyBackbone=MiniGPT-42025.011379375125—
DoLaBackbone=MiniGPT-42025.011379075.33122.67—
InstructBLIP + ICDBackbone=InstructBLIP, Method=ICD2024.03136.679076.67123.33426.67
InstructBLIP (default)Backbone=InstructBLIP, Method=default2024.0313553.3356.6793.33338.33
InstructBLIP + VCDBackbone=InstructBLIP, Method=VCD2024.03123.3381.6755106.67366.67
miniGPT4 + ICDBackbone=miniGPT4, Method=ICD2024.0366.6761.674061.67230.01
miniGPT4 + VCDBackbone=miniGPT4, Method=VCD2024.0348.3331.674045165
miniGPT4 (default)Backbone=miniGPT4, Method=default2024.0346.6726.6738.3338.33150