List your company

Reasoning on Multilingual ARC-E 5 Unseen Languages

95.1ARC-E AccuracyBase
64.00472.07780.1588.223Jun 4, 2026

Evaluation results

MethodLinks
BaseBackbone=Qwen3-8B-Base2026.0695.1
MGDABackbone=Qwen3-8B-Base2026.0694
CAGradBackbone=Qwen3-8B-Base2026.0694
PCGradBackbone=Qwen3-8B-Base2026.0694
V. SFTBackbone=Qwen3-8B-Base2026.0693.4
BaseBackbone=Qwen3-4B-Base2026.0693.3
MGDABackbone=Qwen3-4B-Base2026.0690.8
CAGradBackbone=Qwen3-4B-Base2026.0690.6
PCGradBackbone=Qwen3-4B-Base2026.0690.6
V. SFTBackbone=Qwen3-4B-Base2026.0688.8
CAGradBackbone=Llama-3.1-8B2026.0674.4
PCGradBackbone=Llama-3.1-8B2026.0674.4
MGDABackbone=Llama-3.1-8B2026.0674.1
PCGradBackbone=Meta-Llama-3-8B2026.0673.6
CAGradBackbone=Meta-Llama-3-8B2026.0672.8
MGDABackbone=Meta-Llama-3-8B2026.0672.7
V. SFTBackbone=Llama-3.1-8B2026.0672.6
V. SFTBackbone=Meta-Llama-3-8B2026.0671.5
BaseBackbone=Llama-3.1-8B2026.0670
BaseBackbone=Meta-Llama-3-8B2026.0665.2