Prompt Injection Detection on IHEval Tool-use
0FPRAlignSentinel (Enc-first)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AlignSentinel (Enc-first)LLM=Qwen3-8B, Variant=Enc-first2026.02 | 0 | 0 | |
| AlignSentinel (Enc-first)LLM=Llama3.1-8B, Variant=Enc-first2026.02 | 0 | 0 | |
| AlignSentinel (Enc-first)LLM=Mistral-7B, Variant=Enc-first2026.02 | 0 | 0 | |
| AlignSentinelLLM=Qwen3-8B, Detector Variant=Avg-first2026.02 | 0 | 0 | |
| AlignSentinelLLM=Llama3.1-8B, Detector Variant=Avg-first2026.02 | 0 | 0 | |
| AlignSentinelLLM=Mistral-7B, Detector Variant=Avg-first2026.02 | 0 | 0.03 |