List your company

Object Hallucination on MSCOCO 500 images 2014 (val)

60.6Consistency Score (CS)BeamSearch
12.34424.87237.449.928Nov 29, 2023Apr 5, 2024Aug 12, 2024Dec 19, 2024Apr 26, 2025Sep 2, 2025Jan 9, 2026

Evaluation results

MethodLinks
BeamSearchBase Model=LLaVA-v1.5-7B2026.0160.619.5
VanillaBase Model=LLaVA-v1.5-7B2026.0159.318.2
GreedyBackbone=InstructBLIP, Max New Tokens=5122023.1158.823.7
GreedyBackbone=Shikra, Max New Tokens=5122023.1155.815.4
DoLaBackbone=Shikra, Max New Tokens=5122023.1155.815.1
Beam SearchBackbone=InstructBLIP, Max New Tokens=5122023.1155.615.8
Nucleus SamplingBackbone=Shikra, Max New Tokens=5122023.1155.615.4
Nucleus SamplingBackbone=InstructBLIP, Max New Tokens=5122023.1154.624.8
VCDBase Model=LLaVA-v1.5-7B2026.0152.215.8
Beam SearchBackbone=Shikra, Max New Tokens=5122023.1150.413.3
Nucleus SamplingBackbone=LLaVA-1.5, Max New Tokens=5122023.1148.814.2
Beam SearchBackbone=LLaVA-1.5, Max New Tokens=5122023.1148.813.9
DoLaBackbone=InstructBLIP, Max New Tokens=5122023.1148.415.9
DoLaBackbone=LLaVA-1.5, Max New Tokens=5122023.1147.813.8
PAIBase Model=LLaVA-v1.5-7B2026.0146.714.4
OPERABackbone=InstructBLIP, Max New Tokens=5122023.1146.414.2
GreedyBackbone=LLaVA-1.5, Max New Tokens=5122023.114514.7
VIB-ProbeBase Model=LLaVA-v1.5-7B2026.0144.914.1
OPERABackbone=LLaVA-1.5, Max New Tokens=5122023.1144.612.8
VanillaBase Model=LLaVA-v1.6-7B2026.0140.711.8
BeamSearchBase Model=LLaVA-v1.6-7B2026.0139.210.9
VCDBase Model=LLaVA-v1.6-7B2026.0136.49
OPERABackbone=Shikra, Max New Tokens=5122023.1136.212.1
PAIBase Model=LLaVA-v1.6-7B2026.0135.39.2
Nucleus SamplingBackbone=MiniGPT-4, Max New Tokens=5122023.1132.610.7
DoLaBackbone=MiniGPT-4, Max New Tokens=5122023.1132.210
VIB-ProbeBase Model=LLaVA-v1.6-7B2026.0132.18.7
GreedyBackbone=MiniGPT-4, Max New Tokens=5122023.1131.89.9
Beam SearchBackbone=MiniGPT-4, Max New Tokens=5122023.1130.69.5
NucleusModel=InstructBLIP, Max new tokens=642023.1130.415.7
GreedyModel=InstructBLIP, Max new tokens=642023.113014.5
OPERABackbone=MiniGPT-4, Max New Tokens=5122023.1126.29.5
NucleusModel=LLaVA-1.5, Max new tokens=642023.1126.28.5
GreedyModel=MiniGPT-4, Max new tokens=642023.1124.28.2
DoLaModel=MiniGPT-4, Max new tokens=642023.1124.28.2
NucleusModel=MiniGPT-4, Max new tokens=642023.1123.68.3
Beam SearchModel=MiniGPT-4, Max new tokens=642023.1123.67.8
OPERAModel=MiniGPT-4, Max new tokens=642023.1122.68.2
NucleusModel=Shikra, Max new tokens=642023.1122.67.6
DoLaModel=InstructBLIP, Max new tokens=642023.1122.27.1
GreedyModel=Shikra, Max new tokens=642023.11227
Beam SearchModel=InstructBLIP, Max new tokens=642023.1121.47.2
GreedyModel=LLaVA-1.5, Max new tokens=642023.1120.66.2
DoLaModel=LLaVA-1.5, Max new tokens=642023.1120.46.3
Beam SearchModel=Shikra, Max new tokens=642023.1120.26.4
DoLaModel=Shikra, Max new tokens=642023.1120.26.3
Beam SearchModel=LLaVA-1.5, Max new tokens=642023.1118.85.9
OPERAModel=InstructBLIP, Max new tokens=642023.1116.66.8
OPERAModel=LLaVA-1.5, Max new tokens=642023.1114.25.2
OPERAModel=Shikra, Max new tokens=642023.1114.25.9