List your company

Multilingual Language Understanding on M-MMLU (test)

59.6Overall AccuracyCOMPASS
37.96843.58449.254.816May 31, 2025Jul 24, 2025Sep 16, 2025Nov 10, 2025Jan 3, 2026Feb 26, 2026Apr 22, 2026

Evaluation results

MethodLinks
COMPASSBackbone=Qwen2.5-7B-Instruct2026.0459.6————————
COMPASS-FFTBackbone=Qwen2.5-7B-Instruct, Fine-tuning=Full2026.0458.1————————
LangRankBackbone=Qwen2.5-7B-Instruct2026.0456.7————————
CAUSAL-MULTIBackbone=LLaMa 3.22025.0556.561.258.957.652.349.658.158.155.8
LangSimBackbone=Qwen2.5-7B-Instruct2026.0456.1————————
COMPASS-FFTBackbone=Llama-3.1-Instruct-8B, Fine-tuning=Full2026.0455.5————————
COMPASSBackbone=Llama-3.1-Instruct-8B2026.0455.2————————
RandomBackbone=Qwen2.5-7B-Instruct2026.0455.1————————
CAUSAL-MULTIBackbone=Phi 42025.0554.757.464.760.958.448.244.150.653.1
TargetBackbone=Qwen2.5-7B-Instruct2026.0454.6————————
REFLECTBackbone=Phi 42025.0553.165.260.35757.845.5454845.7
Pretrained (ZS)Backbone=Qwen2.5-7B-Instruct, Evaluation Protocol=Zero-shot2026.0452.9————————
COMPASSBackbone=Phi-4-Mini-Instruct-3.8B2026.0452.4————————
CAUSAL-NATIVEBackbone=LLaMa 3.22025.0552.257.447.453.653.245.5485854.2
LangRankBackbone=Llama-3.1-Instruct-8B2026.0451.8————————
CONFLICTBackbone=Phi 42025.0551.456.852.645.743.547.650.452.752.7
MULTI-RELATEDBackbone=Phi 42025.0551.449.753.256.345.549.744.258.154.3
REFLECTBackbone=LLaMa 3.22025.0551.241.951.95754.653.150.44852.7
CAUSAL-NATIVEBackbone=Phi 42025.0551.251.660.858.958.444.14144.550
LangSimBackbone=Llama-3.1-Instruct-8B2026.0451————————
TargetBackbone=Llama-3.1-Instruct-8B2026.0450.8————————
RandomBackbone=Llama-3.1-Instruct-8B2026.0450.6————————
MOREINFOBackbone=Phi 42025.0550.548.446.854.348.740.754.360.150.4
COMPASS-FFTBackbone=Phi-4-Mini-Instruct-3.8B, Fine-tuning=Full2026.0449.9————————
CONFLICTBackbone=LLaMa 3.22025.0549.65153.253.649.442.152.751.443.4
Pretrained (ZS)Backbone=Llama-3.1-Instruct-8B, Evaluation Protocol=Zero-shot2026.0449.1————————
ASK CALIBackbone=LLaMa 3.22025.0548.564.644.269.566.226.938.830.447.3
MOREINFOBackbone=LLaMa 3.22025.0548.453.650.64947.449.73851.447.3
AllBackbone=Qwen2.5-7B-Instruct2026.0448————————
MULTI-RELATEDBackbone=LLaMa 3.22025.0547.949.753.247.748.147.646.64644.2
LangRankBackbone=Phi-4-Mini-Instruct-3.8B2026.0447.3————————
LangSimBackbone=Phi-4-Mini-Instruct-3.8B2026.0447.3————————
ASK CALIBackbone=Phi 42025.0544.938.142.34943.544.244.251.446.7
RandomBackbone=Phi-4-Mini-Instruct-3.8B2026.0444.9————————
TargetBackbone=Phi-4-Mini-Instruct-3.8B2026.0444.7————————
AllBackbone=Llama-3.1-Instruct-8B2026.0443.8————————
Pretrained (ZS)Backbone=Phi-4-Mini-Instruct-3.8B, Evaluation Protocol=Zero-shot2026.0443.5————————
AllBackbone=Phi-4-Mini-Instruct-3.8B2026.0438.8————————