List your company
ResearchFormerly WizWand

General Reasoning on ARC Challenge (Capability)

77.3CapabilityCSULoRA
67.898470.339272.7875.2208May 28, 2026
Updated 5mo ago

Evaluation results

MethodLinks
CSULoRAModel=Gemma-3-4B-it2026.0577.3
Base modelModel=Gemma-3-4B-it2026.0577.13
SafeLoRAModel=Gemma-3-4B-it2026.0577.13
SPLoRAModel=Gemma-3-4B-it2026.0576.28
LoRAModel=Gemma-3-4B-it2026.0576.11
SaLoRAModel=Gemma-3-4B-it2026.0575.85
AlignGuardModel=Gemma-3-4B-it2026.0575.77
LoRAModel=Llama-3.2-3B-Instruct2026.0573.38
SafeLoRAModel=Llama-3.2-3B-Instruct2026.0573.38
SPLoRAModel=Llama-3.2-3B-Instruct2026.0573.29
CSULoRAModel=Llama-3.2-3B-Instruct2026.0573.29
AlignGuardModel=Llama-3.2-3B-Instruct2026.0572.78
SaLoRAModel=Llama-3.2-3B-Instruct2026.0572.35
Base modelModel=Llama-3.2-3B-Instruct2026.0572.01
RESTAModel=Llama-3.2-3B-Instruct2026.0571.16
RESTAModel=Gemma-3-4B-it2026.0568.26