List your company

Object Hallucination Evaluation on MSCOCO 2014 (val)

56.8CHAIRsVDD-None
19.98429.54239.148.658Dec 18, 2024Mar 15, 2025Jun 11, 2025Sep 7, 2025Dec 3, 2025Mar 1, 2026May 28, 2026

Evaluation results

MethodLinks
VDD-NoneBackbone=LLaVA-1.5-7B, Regime=Greedy2026.0556.816.57—81.84
BaselineBackbone=LLaVA-1.5-7B, Regime=Nucleus2026.0556.617.84—73.13
VDD-NoneBackbone=LLaVA-1.5-7B, Regime=Nucleus2026.0556.617.04—76.04
VanillaDecoding=Greedy, Model=Shikra2025.1254.615——
BeamBackbone=LLaVA-1.5, Number of beams=52024.1253.8415.687.47—
VanillaDecoding=Beam Search, Model=Shikra2025.1253.414.1——
VCDBackbone=LLaVA-1.52024.1251.9215.4283.12—
DoLaBackbone=LLaVA-1.52024.1250.8814.6482.41—
BaselineBackbone=LLaVA-1.5-7B, Regime=Beam2026.0550.814.97—77.4
CODEBackbone=InstructBLIP2024.1250.7614.1288.57—
VCDBackbone=InstructBLIP2024.1250.7214.4290.39—
VCDDecoding=Beam Search, Model=LLAVA-1.52025.125014.4——
VanillaDecoding=Beam Search, Model=LLAVA-1.52025.1249.814——
GreedyBackbone=LLaVA-1.52024.1249.6814.3283.06—
BaselineBackbone=Qwen-VL-Chat, Regime=Nucleus2026.0549.413.07—73.6
VCDBackbone=LLaVA-1.5-7B, Regime=Nucleus2026.0549.214.79—75.34
PAIBackbone=Qwen-VL-Chat, Regime=Greedy2026.054912.54—76.56
BeamBackbone=InstructBLIP, Number of beams=52024.1248.5613.594.87—
PAIBackbone=LLaVA-1.5-7B, Regime=Beam2026.0548.214.75—78.84
CODEBackbone=LLaVA-1.52024.1247.9614.2678.52—
BaselineBackbone=LLaVA-1.5-7B, Regime=Greedy2026.0547.813.03—78.94
VCDBackbone=Qwen-VL-Chat, Regime=Nucleus2026.0547.612.06—74.89
BRACSBackbone=LLaVA-1.5-7B, Regime=Nucleus2026.0547.213.6—74.85
SPINBackbone=LLaVA-1.5-7B, Regime=Beam2026.0546.614.2—73.54
BaselineBackbone=Qwen-VL-Chat, Regime=Greedy2026.0546.612.52—75.6
EAHBackbone=InstructBLIP2024.1246.413.1392.33—
DoLaBackbone=InstructBLIP2024.12461390.75—
OPERABackbone=InstructBLIP, Number of beams=52024.1245.7613.0692.46—
BaselineBackbone=Qwen-VL-Chat, Regime=Beam2026.0545.512.97—76.64
GreedyBackbone=InstructBLIP2024.1245.3212.9891.06—
BRACSBackbone=LLaVA-1.5-7B, Regime=Beam2026.0544.813.52—76.66
VDD-NoneBackbone=Qwen-VL-Chat, Regime=Nucleus2026.0544.411.55—76.11
PAIBackbone=LLaVA-1.5-7B, Regime=Greedy2026.0544.312.58—75.03
OPERABackbone=LLaVA-1.5, Number of beams=52024.1244.2813.3675.88—
VanillaDecoding=Greedy, Model=LLAVA-1.52025.1243.813——
VCDDecoding=Greedy, Model=LLAVA-1.52025.1243.513.8——
SPINBackbone=LLaVA-1.5-7B, Regime=Greedy2026.0543.214.42—73.65
BRACSBackbone=Qwen-VL-Chat, Regime=Nucleus2026.0542.211.02—72.05
VDD-NoneBackbone=Qwen-VL-Chat, Regime=Greedy2026.0541.410.43—72.76
SPINBackbone=Qwen-VL-Chat, Regime=Beam2026.054111.53—75.84
BRACSBackbone=Qwen-VL-Chat, Regime=Greedy2026.0540.811.17—77.4
PAIBackbone=Qwen-VL-Chat, Regime=Beam2026.0540.211.08—75.36
BRACSBackbone=LLaVA-1.5-7B, Regime=Greedy2026.054010.88—77.33
EAHBackbone=LLaVA-1.52024.1238.7611.0586.28—
BRACSBackbone=Qwen-VL-Chat, Regime=Beam2026.053810.48—75.24
VHRBackbone=InstructBLIP2024.1237.769.75106.49—
PAIDecoding=Beam Search, Model=Shikra2025.1236.29.8——
VanillaDecoding=Beam Search, Model=MiniGPT-42025.1234.610.1——
VanillaDecoding=Greedy, Model=MiniGPT-42025.1233.810.4——
VHRBackbone=LLaVA-1.52024.1233.329.7181.33—
VISTADecoding=Beam Search, Model=Shikra2025.12339.6——
PAIDecoding=Beam Search, Model=MiniGPT-42025.1231.89.9——
PAIDecoding=Greedy, Model=Shikra2025.1231.68.9——
VISTADecoding=Greedy, Model=Shikra2025.1231.48.6——
SPINBackbone=Qwen-VL-Chat, Regime=Greedy2026.05318.68—72.38
VCDBackbone=LLaVA-NeXT2024.1230.88.72157.72—
[15]Decoding=Beam Search, Model=MiniGPT-42025.1229.88.8——
[15]Decoding=Greedy, Model=LLAVA-1.52025.1229.58.8——
[15]Decoding=Beam Search, Model=LLAVA-1.52025.1229.48.5——
GreedyBackbone=LLaVA-NeXT2024.1229.088.08157.06—
PAIDecoding=Greedy, Model=LLAVA-1.52025.1228.87.8——
DoLaBackbone=LLaVA-NeXT2024.1228.768.12155.75—
EAHBackbone=LLaVA-NeXT2024.1228.136.62142.75—
CODEBackbone=LLaVA-NeXT2024.1227.847.98151.51—
PAIDecoding=Beam Search, Model=LLAVA-1.52025.1227.57.8——
VISTADecoding=Greedy, Model=LLAVA-1.52025.1227.27.3——
BeamBackbone=LLaVA-NeXT, Number of beams=52024.1225.726.92160.64—
VHRBackbone=LLaVA-NeXT2024.1224.966.8156.92—
VEGASDecoding=Greedy, Model=LLAVA-1.52025.1224.87.1——
VEGASDecoding=Beam Search, Model=Shikra2025.1224.58.8——
PAIDecoding=Greedy, Model=MiniGPT-42025.12249——
VEGASDecoding=Greedy, Model=Shikra2025.12248.1——
VISTADecoding=Beam Search, Model=LLAVA-1.52025.12247.5——
[15]Decoding=Beam Search, Model=Shikra2025.1223.59.1——
[15]Decoding=Greedy, Model=MiniGPT-42025.1223.48.8——
[15]Decoding=Greedy, Model=Shikra2025.1223.48.1——
VEGASDecoding=Beam Search, Model=LLAVA-1.52025.1223.27——
VISTADecoding=Beam Search, Model=MiniGPT-42025.1222.88.2——
VISTADecoding=Greedy, Model=MiniGPT-42025.1222.58.7——
VEGASDecoding=Beam Search, Model=MiniGPT-42025.1221.78——
VEGASDecoding=Greedy, Model=MiniGPT-42025.1221.48.4——