Refusal Detection on Do-Not-Answer Portuguese (test)
100Accuracygov.pt baseline
Evaluation Results
| Method | Links | |
|---|---|---|
| gov.pt baseline2026.03 | 100 | |
| Gervásio 70BRAG=true, Model Size=70B2026.03 | 98 | |
| Gervásio 8BRAG=true, Model Size=8B2026.03 | 97 | |
| Llama 3.3 70BModel Size=70B2026.03 | 92 | |
| Qwen 32BModel Size=32B2026.03 | 90 | |
| Gervásio 70BRAG=false, Model Size=70B2026.03 | 87 | |
| Mistral 24BModel Size=24B2026.03 | 86 | |
| Gervásio 8BRAG=false, Model Size=8B2026.03 | 84 | |
| Llama 3.1 8BModel Size=8B2026.03 | 84 |