Explainability classification on Toxic-Chat 0124 (test)
65.99Unsafe F1LEG large
Evaluation results
| Method | Links | |
|---|---|---|
| LEG largeTrain Dataset=Toxic-Chat0124, Model Size=304M2026.01 | 65.99 | |
| LEG largeTrain Dataset=AEGIS 2.0, Model Size=304M2026.01 | 63.18 | |
| LEG largeTrain Dataset=Wild-GuardMix, Model Size=304M2026.01 | 61.56 | |
| LEG baseTrain Dataset=Toxic-Chat0124, Model Size=86M2026.01 | 60.62 | |
| LEG baseTrain Dataset=AEGIS 2.0, Model Size=86M2026.01 | 59.78 | |
| LEG baseTrain Dataset=Wild-GuardMix, Model Size=86M2026.01 | 58.86 | |
| Word Baseline largeTrain Dataset=AEGIS 2.0, Model Size=304M2026.01 | 57.22 | |
| Word Baseline largeTrain Dataset=Wild-GuardMix, Model Size=304M2026.01 | 55.45 | |
| LEG xsTrain Dataset=AEGIS 2.0, Model Size=22M2026.01 | 51.19 | |
| GPT-4o-minizero-shot=true2026.01 | 50.96 | |
| Word Baseline baseTrain Dataset=AEGIS 2.0, Model Size=86M2026.01 | 50.74 | |
| Word Baseline baseTrain Dataset=Wild-GuardMix, Model Size=86M2026.01 | 50.5 | |
| LEG largesize=large2026.01 | 50.36 | |
| LEG xsTrain Dataset=Wild-GuardMix, Model Size=22M2026.01 | 48.77 | |
| Word Baseline largeTrain Dataset=Toxic-Chat0124, Model Size=304M2026.01 | 45.49 | |
| LEG xsTrain Dataset=Toxic-Chat0124, Model Size=22M2026.01 | 44.63 | |
| Word Baseline baseTrain Dataset=Toxic-Chat0124, Model Size=86M2026.01 | 38.62 | |
| LEG basesize=base2026.01 | 36.12 | |
| SHAP Baseline largeTrain Dataset=Toxic-Chat0124, Model Size=-2026.01 | 21.47 | |
| SHAP Baseline largeTrain Dataset=AEGIS 2.0, Model Size=-2026.01 | 20.94 | |
| LIME Baseline largeTrain Dataset=Toxic-Chat0124, Model Size=-2026.01 | 17.69 | |
| SHAP Baseline largeTrain Dataset=Wild-GuardMix, Model Size=-2026.01 | 15.76 | |
| LIME Baseline baseTrain Dataset=Toxic-Chat0124, Model Size=-2026.01 | 13.63 | |
| SHAP Baseline baseTrain Dataset=Toxic-Chat0124, Model Size=-2026.01 | 11.78 | |
| LIME Baseline largeTrain Dataset=Wild-GuardMix, Model Size=-2026.01 | 10.03 | |
| LIME Baseline largeTrain Dataset=AEGIS 2.0, Model Size=-2026.01 | 9.42 | |
| SHAP Baseline baseTrain Dataset=AEGIS 2.0, Model Size=-2026.01 | 9.03 | |
| LIME Baseline baseTrain Dataset=Wild-GuardMix, Model Size=-2026.01 | 8.4 | |
| SHAP Baseline baseTrain Dataset=Wild-GuardMix, Model Size=-2026.01 | 7.28 | |
| LIME Baseline baseTrain Dataset=AEGIS 2.0, Model Size=-2026.01 | 6.3 |