List your company

Multilingual Language Understanding on MMMLU (Language Specific Averages)

60.5Accuracy (Korean)Task Arithmetic
42.09246.87151.6556.429Jan 30, 2026

Evaluation results

MethodLinks
Task ArithmeticBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.0160.52237.239.9—
baseBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.0160.317.437.438.4—
Ties MergingBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.0160.324.23740.5—
DAREBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.0160.322.63740—
Layer SwappingBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.016025.336.640.6—
Module-wise SwappingBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.016024.537.740.7—
Layer-wise SwappingBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.0159.924.437.540.6—
Lang-OnlyBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.0159.81937.338.7—
Safety-OnlyBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.0159.725.137.240.7—
Mixed SFTBackbone=Qwen 2 7B, Evaluation Protocol=0-shot2026.0159.620.337.339.1—
baseBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0144.528.938.637.3—
DAREBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0144.426.137.636—
Layer SwappingBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0143.625.337.335.4—
Module-wise SwappingBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0143.627.537.436.2—
Task ArithmeticBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0143.526.137.535.7—
Mixed SFTBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0143.424.637.535.2—
Ties MergingBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0143.324.837.235.1—
Lang-OnlyBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0143.124.738.235.3—
Layer-wise SwappingBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0142.927.637.536—
Safety-OnlyBackbone=LLaMA 3.1 8B it, Evaluation Protocol=0-shot2026.0142.824.336.734.6—
Gemma 3 27BModel size=27B, Thinking mode=false2026.07————70.7
Gemma 4 12BModel size=12B, Thinking mode=true2026.07————83.4
Gemma 4 26B-A4BModel size=26B-A4B, Thinking mode=true2026.07————86.3
Gemma 4 31BModel size=31B, Thinking mode=true2026.07————88.4
Gemma 4 E2BModel size=E2B, Thinking mode=true2026.07————67.4
Gemma 4 E4BModel size=E4B, Thinking mode=true2026.07————76.6