Date Understanding on FLORES-200 10 low-resourced languages
21.2Performance Score (kaz_Cyrl)English Pivot Thought
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| English Pivot ThoughtBase Model=Llama-3.2, English Pivoting (EP)=true, Chain-of-thought (CT)=true2024.11 | 21.2 | 20.4 | 17.2 | 24.4 | 25.6 | 18.8 | 18.8 | 18.8 | 22.4 | 24.4 | 22.16 | |
| English PivotingBase Model=Llama-3.2, English Pivoting (EP)=true2024.11 | 20.8 | 23.2 | 18.8 | 24.4 | 25.6 | 18.4 | 24.4 | 20 | 20 | 22.8 | 21.86 | |
| DIP w/ EP w/o CTBase Model=Llama-3.2, English Pivoting (EP)=true, Chain-of-thought (CT)=false2024.11 | 20.4 | 23.6 | 22.8 | 19.2 | 20.8 | 16 | 18.4 | 15.2 | 21.2 | 19.6 | 21.7 | |
| DIPBase Model=Llama-3.22024.11 | 20.4 | 26.8 | 20 | 29.2 | 24 | 22 | 19.2 | 22.8 | 23.6 | 20.4 | 23.94 | |
| Standard PromptingBase Model=Llama-3.2, Prompting Strategy=Standard2024.11 | 17.2 | 16 | 22.8 | 19.2 | 22 | 20.4 | 11.2 | 21.6 | 13.2 | 12.8 | 17.22 | |
| Non-insertion PromptingBase Model=Llama-3.2, Prompting Strategy=Non-insertion2024.11 | 16 | 14.4 | 19.2 | 16.8 | 13.2 | 12 | 15.6 | 16.4 | 13.6 | 15.2 | 15.16 | |
| DIP w/o EP w/o CTBase Model=Llama-3.2, English Pivoting (EP)=false, Chain-of-thought (CT)=false2024.11 | 14.4 | 20.4 | 16.8 | 17.2 | 20.4 | 16.8 | 10.4 | 17.2 | 18 | 18.4 | 19.16 |