Question Answering on CoQA alpha = 0.25 (test)
0.2347Empirical Error Rate (EER)SAFER (LLaMA-3.1-8B-Instruct)
Evaluation Results
| Method | Links | |
|---|---|---|
| SAFER (LLaMA-3.1-8B-Instruct)Risk level (beta)=0.05, Backbone=LLaMA-3.1-8B-Instruct, Alpha=0.252025.10 | 0.2347 | |
| SAFER (OpenChat-3.5)Risk level (beta)=0.05, Backbone=OpenChat-3.5, Alpha=0.252025.10 | 0.2373 | |
| SAFER (Qwen2.5-3B-Instruct)Risk level (beta)=0.05, Backbone=Qwen2.5-3B-Instruct, Alpha=0.252025.10 | 0.2625 | |
| SAFER (Qwen2.5-14B-Instruct)Risk level (beta)=0.05, Backbone=Qwen2.5-14B-Instruct, Alpha=0.252025.10 | 0.271 | |
| SAFER (LLaMA-3.1-8B-Instruct)Risk level (beta)=0.10, Backbone=LLaMA-3.1-8B-Instruct, Alpha=0.252025.10 | 0.2743 | |
| SAFER (Qwen2.5-7B-Instruct)Risk level (beta)=0.05, Backbone=Qwen2.5-7B-Instruct, Alpha=0.252025.10 | 0.2761 | |
| SAFER (OpenChat-3.5)Risk level (beta)=0.10, Backbone=OpenChat-3.5, Alpha=0.252025.10 | 0.2794 | |
| SAFER (Qwen2.5-3B-Instruct)Risk level (beta)=0.10, Backbone=Qwen2.5-3B-Instruct, Alpha=0.252025.10 | 0.3001 | |
| SAFER (Qwen2.5-14B-Instruct)Risk level (beta)=0.10, Backbone=Qwen2.5-14B-Instruct, Alpha=0.252025.10 | 0.3076 | |
| SAFER (LLaMA-3.1-8B-Instruct)Risk level (beta)=0.15, Backbone=LLaMA-3.1-8B-Instruct, Alpha=0.252025.10 | 0.3139 | |
| SAFER (Qwen2.5-7B-Instruct)Risk level (beta)=0.10, Backbone=Qwen2.5-7B-Instruct, Alpha=0.252025.10 | 0.3145 | |
| SAFER (OpenChat-3.5)Risk level (beta)=0.15, Backbone=OpenChat-3.5, Alpha=0.252025.10 | 0.3201 | |
| SAFER (Qwen2.5-3B-Instruct)Risk level (beta)=0.15, Backbone=Qwen2.5-3B-Instruct, Alpha=0.252025.10 | 0.3389 | |
| SAFER (Qwen2.5-14B-Instruct)Risk level (beta)=0.15, Backbone=Qwen2.5-14B-Instruct, Alpha=0.252025.10 | 0.3469 | |
| SAFER (Qwen2.5-7B-Instruct)Risk level (beta)=0.15, Backbone=Qwen2.5-7B-Instruct, Alpha=0.252025.10 | 0.3521 | |
| SAFER (LLaMA-3.1-8B-Instruct)Risk level (beta)=0.20, Backbone=LLaMA-3.1-8B-Instruct, Alpha=0.252025.10 | 0.3546 | |
| SAFER (OpenChat-3.5)Risk level (beta)=0.20, Backbone=OpenChat-3.5, Alpha=0.252025.10 | 0.3598 | |
| SAFER (Qwen2.5-3B-Instruct)Risk level (beta)=0.20, Backbone=Qwen2.5-3B-Instruct, Alpha=0.252025.10 | 0.3778 | |
| SAFER (Qwen2.5-14B-Instruct)Risk level (beta)=0.20, Backbone=Qwen2.5-14B-Instruct, Alpha=0.252025.10 | 0.3851 | |
| SAFER (Qwen2.5-7B-Instruct)Risk level (beta)=0.20, Backbone=Qwen2.5-7B-Instruct, Alpha=0.252025.10 | 0.3912 | |
| SAFER (LLaMA-3.1-8B-Instruct)Risk level (beta)=0.25, Backbone=LLaMA-3.1-8B-Instruct, Alpha=0.252025.10 | 0.3954 | |
| SAFER (OpenChat-3.5)Risk level (beta)=0.25, Backbone=OpenChat-3.5, Alpha=0.252025.10 | 0.401 | |
| SAFER (Qwen2.5-3B-Instruct)Risk level (beta)=0.25, Backbone=Qwen2.5-3B-Instruct, Alpha=0.252025.10 | 0.4159 | |
| SAFER (Qwen2.5-14B-Instruct)Risk level (beta)=0.25, Backbone=Qwen2.5-14B-Instruct, Alpha=0.252025.10 | 0.4245 | |
| SAFER (Qwen2.5-7B-Instruct)Risk level (beta)=0.25, Backbone=Qwen2.5-7B-Instruct, Alpha=0.252025.10 | 0.4276 | |
| SAFER (LLaMA-3.1-8B-Instruct)Risk level (beta)=0.30, Backbone=LLaMA-3.1-8B-Instruct, Alpha=0.252025.10 | 0.4347 | |
| SAFER (OpenChat-3.5)Risk level (beta)=0.30, Backbone=OpenChat-3.5, Alpha=0.252025.10 | 0.4401 | |
| SAFER (Qwen2.5-3B-Instruct)Risk level (beta)=0.30, Backbone=Qwen2.5-3B-Instruct, Alpha=0.252025.10 | 0.4549 | |
| SAFER (Qwen2.5-14B-Instruct)Risk level (beta)=0.30, Backbone=Qwen2.5-14B-Instruct, Alpha=0.252025.10 | 0.4629 | |
| SAFER (Qwen2.5-7B-Instruct)Risk level (beta)=0.30, Backbone=Qwen2.5-7B-Instruct, Alpha=0.252025.10 | 0.464 | |
| SAFER (LLaMA-3.1-8B-Instruct)Risk level (beta)=0.35, Backbone=LLaMA-3.1-8B-Instruct, Alpha=0.252025.10 | 0.4729 | |
| SAFER (OpenChat-3.5)Risk level (beta)=0.35, Backbone=OpenChat-3.5, Alpha=0.252025.10 | 0.4802 | |
| SAFER (Qwen2.5-3B-Instruct)Risk level (beta)=0.35, Backbone=Qwen2.5-3B-Instruct, Alpha=0.252025.10 | 0.4932 | |
| SAFER (Qwen2.5-7B-Instruct)Risk level (beta)=0.35, Backbone=Qwen2.5-7B-Instruct, Alpha=0.252025.10 | 0.5015 | |
| SAFER (Qwen2.5-14B-Instruct)Risk level (beta)=0.35, Backbone=Qwen2.5-14B-Instruct, Alpha=0.252025.10 | 0.5023 | |
| SAFER (LLaMA-3.1-8B-Instruct)Risk level (beta)=0.40, Backbone=LLaMA-3.1-8B-Instruct, Alpha=0.252025.10 | 0.5142 | |
| SAFER (OpenChat-3.5)Risk level (beta)=0.40, Backbone=OpenChat-3.5, Alpha=0.252025.10 | 0.5203 | |
| SAFER (Qwen2.5-3B-Instruct)Risk level (beta)=0.40, Backbone=Qwen2.5-3B-Instruct, Alpha=0.252025.10 | 0.5324 | |
| SAFER (Qwen2.5-7B-Instruct)Risk level (beta)=0.40, Backbone=Qwen2.5-7B-Instruct, Alpha=0.252025.10 | 0.5383 | |
| SAFER (Qwen2.5-14B-Instruct)Risk level (beta)=0.40, Backbone=Qwen2.5-14B-Instruct, Alpha=0.252025.10 | 0.5395 |