Language Modeling on PTB (Adversarial Robustness)
28.32Perplexity (Before Attack)GDF-BFA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GDF-BFABackbone=Llama3-8B, # of bit flips=1, Quantization=INT82025.11 | 28.32 | 32,048.31 | |
| Llama3-8BModel=Llama3-8B, # of bit flips=1, Quantization precision=4-bit, Attack=GDF-BFA2025.11 | 31.59 | 366,773.58 | |
| GDF-BFABackbone=Qwen3-8B, # of bit flips=6, Quantization=INT82025.11 | 42.52 | 624.84 | |
| GDF-BFABackbone=Llama2-7B, # of bit flips=1, Quantization=INT82025.11 | 47.34 | 2,395.26 | |
| GDF-BFABackbone=Mistral-7B-v0.3, # of bit flips=1, Quantization=INT82025.11 | 48.94 | 4,074.4 | |
| Llama2-7BModel=Llama2-7B, # of bit flips=1, Quantization precision=4-bit, Attack=GDF-BFA2025.11 | 53.751 | 6,717.69 | |
| GDF-BFABackbone=Llama2-13B, # of bit flips=3, Quantization=INT82025.11 | 57.66 | 909.14 | |
| Llama2-13BModel=Llama2-13B, # of bit flips=3, Quantization precision=4-bit, Attack=GDF-BFA2025.11 | 58.11 | 123.03 |