Harmfulness Evaluation on I-CoNa
0Harmfulness RateNLCf/800 step
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| NLCf/800 stepSteps=800, Loss=NLCf2026.05 | 0 | 2.5 | |
| EMD/800 stepSteps=800, Loss=EMD2026.05 | 0 | 2.5 | |
| Mistral MoE-XLArchitecture=MoE-XL2026.05 | 1 | 0.01 | |
| Qwen-3B MoE-XLArchitecture=MoE-XL2026.05 | 25 | 0.02 | |
| Mistral2026.05 | 40 | 3 | |
| Qwen-3B2026.05 | 52 | 0.04 |