Answerability Prediction on CODE n=30 (matched pairs)
85AUCRefusal (Keyword Classifier)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Refusal (Keyword Classifier)Model=Qwen2026.05 | 85 | 83 | |
| Geometry (own_dist)Model=Qwen2026.05 | 81.8 | 73.3 | |
| Geometry (own_dist)Model=Mistral2026.05 | 79.6 | 68.9 | |
| Geometry (own_dist)Model=Llama2026.05 | 77.4 | 75.8 | |
| Refusal (Keyword Classifier)Model=Mistral2026.05 | 73.3 | 63.6 | |
| Refusal (Keyword Classifier)Model=Llama2026.05 | 63.3 | 42.1 | |
| SC (Self-Consistency)Model=Mistral2026.05 | 49.7 | — | |
| SC (Self-Consistency)Model=Llama2026.05 | 44.1 | — | |
| SC (Self-Consistency)Model=Qwen2026.05 | 36.9 | — |