Natural Language Understanding on EuroEval German
0.1EuroEval Scoreaya-3B (Monolingual Paired DPO)
Evaluation Results
| Method | Links | |
|---|---|---|
| aya-3B (Monolingual Paired DPO)Language model=aya-3B, Regime=Monolingual, Post-training=DPO, Data Selection=Paired2026.05 | 0.1 | |
| aya-3B (Multilingual Paired DPO)Language model=aya-3B, Regime=Multilingual, Post-training=DPO, Data Selection=Paired2026.05 | 0.1 | |
| EuroLLM-9B (Monolingual In-lang SFT)Language model=EuroLLM-9B, Regime=Monolingual, Post-training=SFT, Data Selection=In-lang2026.05 | 0.4 | |
| MAPO (13B)Baseline Model=MAPO, Parameters=13B, Comparison=aya-3B2026.05 | 0.7 | |
| EuroLLM-9B (Multilingual Max-R SFT)Language model=EuroLLM-9B, Regime=Multilingual, Post-training=SFT, Data Selection=Max-R2026.05 | 0.9 | |
| EuroLLM-9B (Monolingual Paired DPO)Language model=EuroLLM-9B, Regime=Monolingual, Post-training=DPO, Data Selection=Paired2026.05 | 1 | |
| EuroLLM-9B (Multilingual Paired DPO)Language model=EuroLLM-9B, Regime=Multilingual, Post-training=DPO, Data Selection=Paired2026.05 | 1.2 | |
| ICR (8B)Baseline Model=ICR, Parameters=8B, Comparison=aya-3B2026.05 | 1.7 | |
| EuroLLM-9B (Monolingual Max-R SFT)Language model=EuroLLM-9B, Regime=Monolingual, Post-training=SFT, Data Selection=Max-R2026.05 | 2 | |
| EuroLLM-9B (Multilingual All Lang SFT)Language model=EuroLLM-9B, Regime=Multilingual, Post-training=SFT, Data Selection=All Lang2026.05 | 3.8 | |
| aya-3B (Monolingual In-lang SFT)Language model=aya-3B, Regime=Monolingual, Post-training=SFT, Data Selection=In-lang2026.05 | 4.2 | |
| aya-3B (Monolingual Max-R SFT)Language model=aya-3B, Regime=Monolingual, Post-training=SFT, Data Selection=Max-R2026.05 | 5.7 | |
| MAPO (13B)Baseline Model=MAPO, Parameters=13B, Comparison=EuroLLM-9B2026.05 | 5.7 | |
| ICR (8B)Baseline Model=ICR, Parameters=8B, Comparison=EuroLLM-9B2026.05 | 6.7 | |
| aya-3B (Multilingual All Lang SFT)Language model=aya-3B, Regime=Multilingual, Post-training=SFT, Data Selection=All Lang2026.05 | 6.9 | |
| aya-3B (Multilingual Max-R SFT)Language model=aya-3B, Regime=Multilingual, Post-training=SFT, Data Selection=Max-R2026.05 | 7.6 | |
| aya-3B (Base)Language model=aya-3B, Regime=Baseline2026.05 | 42.4 | |
| EuroLLM-9B (Base)Language model=EuroLLM-9B, Regime=Baseline2026.05 | 47.4 |