Mathematical Reasoning on GSM8K (ASR & Pass@1)
76.7ASRSFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SFTPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.12 | 76.7 | 22.4 | |
| SafeLoRAPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.12 | 24.4 | 23 | |
| A-GEMPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.12 | 16.6 | 20.6 | |
| SafeInstrPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.12 | 5.8 | 21.4 | |
| LwFPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.12 | 2.7 | 20.4 | |
| DERPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.12 | 2.3 | 20.8 | |
| LisaPoisoned fine-tuning=p=0.1, Backbone=Safety-aligned LLaMA2-7B2025.12 | 1.8 | 20 |