List your company
ResearchFormerly WizWand

Reasoning on SIQA (leave-one-out setup)

82.4Average AccuracyFrom scratch
72.62475.16277.780.238Mar 10, 2026
Updated 7mo ago

Evaluation results

MethodLinks
From scratchModel=Gemma-3 4B, Setup=LoRA2026.0382.4
Mashup LearningModel=Gemma-3 4B, Setup=LoRA2026.0382.2
Mashup LearningModel=Gemma-2 2B, Setup=LoRA2026.0381.7
From scratchModel=Gemma-2 2B, Setup=LoRA2026.0381.5
Mashup LearningModel=Gemma-3 4B, Setup=Full FT2026.0381
Mashup LearningModel=Gemma-2 2B, Setup=Full FT2026.0379.5
From scratchModel=Gemma-3 4B, Setup=Full FT2026.0379.5
From scratchModel=Gemma-2 2B, Setup=Full FT2026.0379
From scratchModel=Gemma-3 1B, Setup=LoRA2026.0374.2
Mashup LearningModel=Gemma-3 1B, Setup=LoRA2026.0374.1
Mashup LearningModel=Gemma-3 1B, Setup=Full FT2026.0373.3
From scratchModel=Gemma-3 1B, Setup=Full FT2026.0373