List your company

Hallucination Reduction on CAA Hallucination Benchmark Multiple-Choice

86.65Alignment ProbabilityINNSTEER
37.19850.036562.87575.7135Jun 7, 2026

Evaluation results

MethodLinks
INNSTEERModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0686.652.98
INNSTEERModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0683.324.58
ODEModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0677.63.03
TFModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0675.33.04
ITIModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0674.23.03
LAcTModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0673.83.03
ODEModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0673.54.53
HPRModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0673.13.02
INNSTEERModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0672.413.2
TFModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0671.24.55
ITIModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0670.44.52
LAcTModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.06704.52
HPRModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0669.14.5
CAAModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0669.023
RECModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0668.53
INNSTEERModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0668.283.18
BaseModel=LLaMA3-8B, Steering direction=Negative steering2026.0668.052.94
BaseModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0668.052.94
PoEModel=LLaMA3-8B, Steering direction=Negative steering2026.0667.62.99
PoDModel=LLaMA3-8B, Steering direction=Negative steering2026.0667.42.98
PoEModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0667.42.99
PoDModel Backbone=LLaMA3-8B, Steering Direction=Positive2026.0667.22.98
RECModel=LLaMA3-8B, Steering direction=Negative steering2026.0667.13
CAAModel=LLaMA3-8B, Steering direction=Negative steering2026.0666.93
CAAModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0665.584.48
HPRModel=LLaMA3-8B, Steering direction=Negative steering2026.0665.43.02
RECModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0665.14.49
BaseModel=LLaMA3-3B, Steering direction=Negative steering2026.0664.613.95
BaseModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0664.614.45
ITIModel=LLaMA3-8B, Steering direction=Negative steering2026.0664.53.03
LAcTModel=LLaMA3-8B, Steering direction=Negative steering2026.0664.23.03
ODEModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0664.13.12
PoEModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0664.054.47
TFModel=LLaMA3-8B, Steering direction=Negative steering2026.0663.83.04
PoDModel Backbone=LLaMA3-3B, Steering Direction=Positive2026.0663.84.47
PoEModel=LLaMA3-3B, Steering direction=Negative steering2026.0663.43.97
PoDModel=LLaMA3-3B, Steering direction=Negative steering2026.0663.23.97
ODEModel=LLaMA3-8B, Steering direction=Negative steering2026.0662.53.03
RECModel=LLaMA3-3B, Steering direction=Negative steering2026.0662.43.99
TFModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0662.43.13
ITIModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0661.83.12
CAAModel=LLaMA3-3B, Steering direction=Negative steering2026.0661.783.98
LAcTModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0661.53.11
INNSTEERModel=LLaMA3-8B, Steering direction=Negative steering2026.0661.12.98
INNSTEERModel=LLaMA3-3B, Steering direction=Negative steering2026.0660.794.07
HPRModel=LLaMA3-3B, Steering direction=Negative steering2026.0660.74.01
HPRModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0660.73.1
ITIModel=LLaMA3-3B, Steering direction=Negative steering2026.0659.84.02
LAcTModel=LLaMA3-3B, Steering direction=Negative steering2026.0659.54.02
TFModel=LLaMA3-3B, Steering direction=Negative steering2026.0658.94.05
CAAModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0658.123.08
RECModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0657.93.07
BaseModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0657.363.05
ODEModel=LLaMA3-3B, Steering direction=Negative steering2026.0657.24.03
PoEModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0657.053.05
CAAModel=Qwen2.5-3B, Steering direction=Negative steering2026.0656.93.08
PoDModel Backbone=Qwen2.5-3B, Steering Direction=Positive2026.0656.93.04
RECModel=Qwen2.5-3B, Steering direction=Negative steering2026.0656.83.07
BaseModel=Qwen2.5-3B, Steering direction=Negative steering2026.0656.793.05
PoEModel=Qwen2.5-3B, Steering direction=Negative steering2026.0656.63.05
PoDModel=Qwen2.5-3B, Steering direction=Negative steering2026.0656.53.04
ODEModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0655.93.18
HPRModel=Qwen2.5-3B, Steering direction=Negative steering2026.0655.63.1
ITIModel=Qwen2.5-3B, Steering direction=Negative steering2026.0654.93.12
LAcTModel=Qwen2.5-3B, Steering direction=Negative steering2026.0654.73.11
TFModel=Qwen2.5-3B, Steering direction=Negative steering2026.0654.43.13
TFModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0653.83.19
ODEModel=Qwen2.5-3B, Steering direction=Negative steering2026.0653.63.12
ITIModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0652.93.18
LAcTModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0652.43.17
HPRModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0651.83.16
INNSTEERModel=Qwen2.5-3B, Steering direction=Negative steering2026.0650.273.18
CAAModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0647.593.12
RECModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0647.13.11
INNSTEERModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0646.993.25
BaseModel=Qwen2.5-7B, Steering direction=Negative steering2026.0646.33.1
BaseModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0646.33.1
PoEModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.06463.1
PoDModel Backbone=Qwen2.5-7B, Steering Direction=Positive2026.0645.83.09
PoEModel=Qwen2.5-7B, Steering direction=Negative steering2026.0645.33.1
PoDModel=Qwen2.5-7B, Steering direction=Negative steering2026.0645.13.09
ODEModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0644.83.22
RECModel=Qwen2.5-7B, Steering direction=Negative steering2026.0644.23.11
CAAModel=Qwen2.5-7B, Steering direction=Negative steering2026.0643.923.12
TFModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0643.43.23
HPRModel=Qwen2.5-7B, Steering direction=Negative steering2026.0642.83.18
ITIModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0642.83.22
LAcTModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0642.63.21
HPRModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0642.23.2
ITIModel=Qwen2.5-7B, Steering direction=Negative steering2026.0641.73.18
LAcTModel=Qwen2.5-7B, Steering direction=Negative steering2026.0641.33.17
CAAModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0641.163.18
TFModel=Qwen2.5-7B, Steering direction=Negative steering2026.0640.93.19
RECModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0640.83.17
BaseModel=Qwen2.5-32B, Steering direction=Negative steering2026.0640.423.15
BaseModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0640.423.15
PoEModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0640.13.15
PoDModel Backbone=Qwen2.5-32B, Steering Direction=Positive2026.0639.93.14
ODEModel=Qwen2.5-7B, Steering direction=Negative steering2026.0639.83.18
PoEModel=Qwen2.5-32B, Steering direction=Negative steering2026.0639.13.15