List your company

Reasoning on Multilingual ARC-E 8 Seen Languages

85.7ARC-E AccuracyBase
52.9461.44569.9578.455Jun 4, 2026

Evaluation results

MethodLinks
BaseBackbone=Qwen3-8B-Base2026.0685.7
PCGradBackbone=Qwen3-8B-Base2026.0685.4
MGDABackbone=Qwen3-8B-Base2026.0685.3
CAGradBackbone=Qwen3-8B-Base2026.0685.2
V. SFTBackbone=Qwen3-8B-Base2026.0684.8
BaseBackbone=Qwen3-4B-Base2026.0681.7
CAGradBackbone=Qwen3-4B-Base2026.0680.7
MGDABackbone=Qwen3-4B-Base2026.0680.2
PCGradBackbone=Qwen3-4B-Base2026.0680.1
V. SFTBackbone=Qwen3-4B-Base2026.0678.9
PCGradBackbone=Meta-Llama-3-8B2026.0664.3
CAGradBackbone=Meta-Llama-3-8B2026.0663.9
MGDABackbone=Llama-3.1-8B2026.0663.9
MGDABackbone=Meta-Llama-3-8B2026.0663.7
CAGradBackbone=Llama-3.1-8B2026.0663.7
PCGradBackbone=Llama-3.1-8B2026.0663.5
V. SFTBackbone=Meta-Llama-3-8B2026.0661.5
V. SFTBackbone=Llama-3.1-8B2026.0661.3
BaseBackbone=Llama-3.1-8B2026.0660
BaseBackbone=Meta-Llama-3-8B2026.0654.2