List your company

Hallucination Mitigation on CHAIR

75CHAIR_SMiddleLayer
14.6830.344661.66Feb 4, 2026

Evaluation results

MethodLinks
MiddleLayerBase Model=InstructBLIP2026.027567.2
SPARCBase Model=InstructBLIP2026.027070
OPERABase Model=InstructBLIP2026.026869.2
PAIBase Model=InstructBLIP2026.0263.471.1
VCDBase Model=InstructBLIP2026.0261.671.6
BaselineBase Model=InstructBLIP2026.0261.471.6
VCDBase Model=LLaVA-1.52026.025671.1
KVSmooth (w/o Ada.)Base Model=InstructBLIP2026.0247.874.1
SPARCBase Model=LLaVA-1.52026.0245.678.9
OPERABase Model=LLaVA-1.52026.0244.278.6
KVSmoothBase Model=InstructBLIP2026.0242.275.1
BaselineBase Model=LLaVA-1.52026.0241.877.5
KVSmooth (w/o Ada.)Base Model=LLaVA-1.52026.0236.279.2
BaselineBase Model=MiniGPT-42026.0231.869.9
VCDBase Model=MiniGPT-42026.023170
OPERABase Model=MiniGPT-42026.0227.469.4
SPARCBase Model=MiniGPT-42026.0226.872.5
PAIBase Model=MiniGPT-42026.0224.671
MiddleLayerBase Model=MiniGPT-42026.0224.671.2
KVSmooth (w/o Ada.)Base Model=MiniGPT-42026.022371.3
PAIBase Model=LLaVA-1.52026.0222.675.5
KVSmoothBase Model=LLaVA-1.52026.0218.279.2
MiddleLayerBase Model=LLaVA-1.52026.0217.875.9
KVSmoothBase Model=MiniGPT-42026.021771.7