List your company

Object Hallucination in Open-ended Captioning on CHAIR (test)

62.3CHAIR_SM3ID
26.4235.73545.0554.365Feb 17, 2026

Evaluation results

MethodLinks
M3IDBackbone=InstructBLIP, Max Token=1282026.0262.318.2————————————————
WoodpeckerBackbone=InstructBLIP, Max Token=1282026.0260.817.6————————————————
VCDBackbone=InstructBLIP, Max Token=1282026.0260.718————————————————
WoodpeckerBackbone=LLaVA-1.5, Max Token=1282026.0257.616.7————————————————
VanillaBackbone=InstructBLIP, Max Token=1282026.0257.417.6————————————————
M3IDBackbone=LLaVA-1.5, Max Token=1282026.0256.615.8————————————————
VanillaBackbone=LLaVA-1.5, Max Token=1282026.0255.116.4————————————————
AGLABackbone=InstructBLIP, Max Token=1282026.0254.816.2————————————————
VCDBackbone=LLaVA-1.5, Max Token=1282026.0254.416.6————————————————
OPERABackbone=InstructBLIP, Max Token=1282026.0254.214.8————————————————
PAIBackbone=InstructBLIP, Max Token=1282026.0254.215.6————————————————
HALCBackbone=InstructBLIP, Max Token=1282026.0253.815.7————————————————
VAFBackbone=InstructBLIP, Max Token=1282026.0253.415.1————————————————
PAIBackbone=LLaVA-1.5, Max Token=1282026.0253.115.1————————————————
AGLABackbone=LLaVA-1.5, Max Token=1282026.0252.414.6————————————————
VARBackbone=InstructBLIP, Max Token=1282026.0252.314.7————————————————
ONLYBackbone=InstructBLIP, Max Token=1282026.0252.215.5————————————————
VanillaBackbone=Qwen-VL, Max Token=1282026.0252.116.7————————————————
PADEBackbone=InstructBLIP, Max Token=1282026.0251.814.2————————————————
WoodpeckerBackbone=Qwen-VL, Max Token=1282026.0251.816.3————————————————
OPERABackbone=LLaVA-1.5, Max Token=1282026.0251.614.2————————————————
HALCBackbone=LLaVA-1.5, Max Token=1282026.025114.8————————————————
VCDBackbone=Qwen-VL, Max Token=1282026.0250.417.2————————————————
VanillaBackbone=LLaVA-1.5-13B, Max Token=1282026.0250.414.7————————————————
VAFBackbone=LLaVA-1.5, Max Token=1282026.0250.114.2————————————————
ONLYBackbone=LLaVA-1.5, Max Token=1282026.0249.814.3————————————————
M3IDBackbone=Qwen-VL, Max Token=1282026.0249.817.4————————————————
AGLABackbone=Qwen-VL, Max Token=1282026.0249.815.6————————————————
VARBackbone=LLaVA-1.5, Max Token=1282026.0249.614.1————————————————
HALCBackbone=Qwen-VL, Max Token=1282026.0249.615.4————————————————
VCDBackbone=LLaVA-1.5-13B, Max Token=1282026.0249.614.3————————————————
PAIBackbone=LLaVA-1.5-13B, Max Token=1282026.0249.414.1————————————————
PAIBackbone=Qwen-VL, Max Token=1282026.0249.115.6————————————————
ONLYBackbone=LLaVA-1.5-13B, Max Token=1282026.0248.913.6————————————————
VAFBackbone=Qwen-VL, Max Token=1282026.0248.714.4————————————————
PADEBackbone=LLaVA-1.5, Max Token=1282026.0248.613.7————————————————
OPERABackbone=Qwen-VL, Max Token=1282026.0248.614.6————————————————
VAFBackbone=LLaVA-1.5-13B, Max Token=1282026.0248.613.4————————————————
VARBackbone=Qwen-VL, Max Token=1282026.0248.413.9————————————————
VARBackbone=LLaVA-1.5-13B, Max Token=1282026.0248.213.2————————————————
ONLYBackbone=Qwen-VL, Max Token=1282026.024814.3————————————————
PADEBackbone=Qwen-VL, Max Token=1282026.0247.813.4————————————————
PADEBackbone=LLaVA-1.5-13B, Max Token=1282026.0247.312.6————————————————
VanillaBackbone=LLaVA-Next, Max Token=1282026.0230.210.9————————————————
VCDBackbone=LLaVA-Next, Max Token=1282026.0229.810.6————————————————
PAIBackbone=LLaVA-Next, Max Token=1282026.0229.410.4————————————————
ONLYBackbone=LLaVA-Next, Max Token=1282026.0228.69.7————————————————
VAFBackbone=LLaVA-Next, Max Token=1282026.0228.49.6————————————————
VARBackbone=LLaVA-Next, Max Token=1282026.0228.29.2————————————————
PADEBackbone=LLaVA-Next, Max Token=1282026.0227.88.9————————————————
CGDType=Training-free mitigation2026.03——52.451.459.15653.852.4545655.457.256.159.852.255.651.654.9
ClipTTTLoRA rank=8, LoRA scaling factor=16, Adaptation iterations=70, Optimizer=AdamW, Learning Rate Scheduler=Cosine annealing2026.03——4844.450.84544.846.648.450.349.64652.248.448.45045.647.9
Greedy DecodingDescription=Standard decoding on corrupted images, Type=Baseline2026.03——56.851.857.854.253.852.655.45555.455.85858.255.45756.455.6
NL-MeansType=Image pre-processing baseline2026.03——57.64253.256.854.2555557.25458.860.857.85555.852.455
PAIType=Training-free mitigation2026.03——58.648.656.654.651.65456.455.656.857.262.258.25455.25755.8
RestormerType=Image pre-processing baseline2026.03——56.247.457.653.653.852.857.25459.456.858.65757.654.255.255.4
VAPType=Training-free mitigation2026.03——5446.856.753.950.753.15250.952.355.255.957.552.649.750.252.8
VCDType=Training-free mitigation2026.03——55.65056.651.454.253.65454.854.657.658.855.85455.253.254.6