Multilingual Question Answering on mGPQA
32.9Accuracyself-cons (ours)
Evaluation Results
| Method | Links | |
|---|---|---|
| self-cons (ours)Model Size=14B2026.05 | 32.9 | |
| S1Model Size=14B2026.05 | 32.1 | |
| LIDRModel Size=14B2026.05 | 29.4 | |
| self-cons (ours)Model Size=7B2026.05 | 28.9 | |
| baseModel Size=14B2026.05 | 28.2 | |
| S1Model Size=7B2026.05 | 25.7 | |
| self-cons (ours)Model Size=1.5B2026.05 | 24.9 | |
| baseModel Size=7B2026.05 | 24.6 | |
| S1Model Size=1.5B2026.05 | 24.5 | |
| LIDRModel Size=7B2026.05 | 23.5 | |
| baseModel Size=1.5B2026.05 | 22.8 | |
| LIDRModel Size=1.5B2026.05 | 18.8 |