Out-of-Distribution Detection on 20NG (ID) vs RTE (OOD)
0.9991AUROCGemma2-2b LR baseline
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Gemma2-2b LR baselineBackbone=Gemma2-2b, Backend=Logistic Regression, Training Protocol=ID-only training2026.05 | 0.9991 | 0.995 | 0.22 | |
| Gemma2-9b LR baselineBackbone=Gemma2-9b, Backend=Logistic Regression, Training Protocol=ID-only training2026.05 | 0.9987 | 0.9994 | 0.3 | |
| Gemma2-9b + 16k SAE + MarkovBackbone=Gemma2-9b, SAE Size=16k, Backend=Markov, Training Protocol=ID-only training2026.05 | 0.9884 | 0.7333 | 4.69 | |
| Gemma2-2b + 16k SAE + MarkovBackbone=Gemma2-2b, SAE Size=16k, Backend=Markov, Training Protocol=ID-only training2026.05 | 0.985 | 0.6422 | 3.25 |