List your company

Object Hallucination Mitigation on MS COCO (500 random samples)

54.4Cs ScoreBaseline
8.53620.44332.3544.257Apr 7, 2026

Evaluation results

MethodLinks
BaselineDecoding=Nucleus, Backbone=Shikra2026.0454.41672.3
BaselineDecoding=Greedy, Backbone=Shikra2026.045315.972.4
PAIDecoding=Nucleus, Backbone=Shikra2026.045014.673.6
PAIDecoding=Greedy, Backbone=Shikra2026.0449.913.974.7
PAIDecoding=Beam, Backbone=Shikra2026.044813.274.9
BaselineDecoding=Beam, Backbone=Shikra2026.0444.213.674.5
OPERADecoding=Beam, Backbone=Shikra2026.0436.212.1—
PAIDecoding=Beam, Backbone=MiniGPT-42026.0431.810.569.2
BaselineDecoding=Beam, Backbone=MiniGPT-42026.0431.610.569.2
BaselineDecoding=Greedy, Backbone=MiniGPT-42026.0430.69.869.4
BaselineDecoding=Nucleus, Backbone=MiniGPT-42026.0430.410.468.7
PAIDecoding=Greedy, Backbone=MiniGPT-42026.0429.39.368.6
PAIDecoding=Nucleus, Backbone=MiniGPT-42026.0428.69.767.2
EAZYDecoding=Greedy, Backbone=Shikra2026.0426.68.9—
OPERADecoding=Beam, Backbone=MiniGPT-42026.0426.29.5—
DIMLDecoding=Greedy, Backbone=Shikra2026.0423.89.472.7
DIMLDecoding=Greedy, Backbone=MiniGPT-42026.0421.4870.8
HaloProbe + BeamDecoding=Beam, Backbone=Shikra2026.0419.65.874.4
HaloProbe + Post-processDecoding=Greedy, Backbone=Shikra2026.0415.6573.4
HaloProbe + Post-processDecoding=Nucleus, Backbone=Shikra2026.0413.24.372.5
HaloProbe + Post-processDecoding=Greedy, Backbone=MiniGPT-42026.0411.84.170.3
HaloProbe + Post-processDecoding=Nucleus, Backbone=MiniGPT-42026.0410.83.768.8
HaloProbe + BeamDecoding=Beam, Backbone=MiniGPT-42026.0410.34.169.7