Prompt Injection Attack Detection on XSum
100TPRDualSentinel
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DualSentinelBackbone=GPT-4o, Attacks=System, Target Form=OnlyTarget, Sequence Length=Long2026.03 | 100 | 0 | |
| DualSentinelBackbone=GPT-4o-mini, Attacks=System, Target Form=OnlyTarget, Sequence Length=Long2026.03 | 100 | 40 | |
| DualSentinelModel=LlaMA2 13B, Attacks=Tree, Target format=OnlyTarget, Target sequence length=Long2026.03 | 100 | 2.8 | |
| DualSentinelModel=LlaMA2 7B, Attacks=Word, Target format=OnlyTarget, Target sequence length=Long2026.03 | 100 | 8 | |
| DualSentinelModel=LlaMA2 7B, Attacks=Phrase, Target format=OnlyTarget, Target sequence length=Long2026.03 | 100 | 7 | |
| DualSentinelModel=Qwen2.5 14B, Attacks=Complete, Target format=OnlyTarget, Target sequence length=Long2026.03 | 100 | 1.6 | |
| DualSentinelModel=Qwen2.5 14B, Attacks=System, Target format=OnlyTarget, Target sequence length=Long2026.03 | 100 | 1.6 | |
| DualSentinelModel=Qwen2.5 7B, Attacks=Word, Target format=OnlyTarget, Target sequence length=Long2026.03 | 100 | 4.82 | |
| DualSentinelModel=Qwen2.5 7B, Attacks=Joint, Target format=OnlyTarget, Target sequence length=Long2026.03 | 100 | 5.71 | |
| DualSentinelBackbone=GPT-4o, Attacks=Average, Target Form=OnlyTarget, Sequence Length=Long2026.03 | 99.8 | 0 | |
| DualSentinelModel=Qwen2.5 14B, Attacks=Ignore, Target format=OnlyTarget, Target sequence length=Long2026.03 | 99.69 | 1.6 | |
| DualSentinelBackbone=GPT-4o, Attacks=Complete, Target Form=OnlyTarget, Sequence Length=Long2026.03 | 99.6 | 0 | |
| DualSentinelModel=LlaMA2 7B, Attacks=Sentence, Target format=OnlyTarget, Target sequence length=Long2026.03 | 99 | 10 | |
| DualSentinelModel=LlaMA2 7B, Attacks=Joint, Target format=OnlyTarget, Target sequence length=Long2026.03 | 99 | 8 | |
| DualSentinelModel=Qwen2.5 7B, Attacks=Phrase, Target format=OnlyTarget, Target sequence length=Long2026.03 | 99 | 7.14 | |
| DualSentinelModel=Qwen2.5 7B, Attacks=Sentence, Target format=OnlyTarget, Target sequence length=Long2026.03 | 99 | 7.77 | |
| DualSentinelBackbone=GPT-4o-mini, Attacks=Average, Target Form=OnlyTarget, Sequence Length=Long2026.03 | 98.6 | 40 | |
| DualSentinelModel=LlaMA2 13B, Attacks=System, Target format=OnlyTarget, Target sequence length=Long2026.03 | 98.3 | 2.8 | |
| DualSentinelModel=Qwen2.5 14B, Attacks=Tree, Target format=OnlyTarget, Target sequence length=Long2026.03 | 97.8 | 1.6 | |
| DualSentinelBackbone=GPT-4o-mini, Attacks=Complete, Target Form=OnlyTarget, Sequence Length=Long2026.03 | 97.2 | 40 | |
| DualSentinelModel=LlaMA2 13B, Attacks=Ignore, Target format=OnlyTarget, Target sequence length=Long2026.03 | 88.55 | 2.8 | |
| DualSentinelModel=LlaMA2 13B, Attacks=Complete, Target format=OnlyTarget, Target sequence length=Long2026.03 | 87.5 | 2.8 |