Agent Memory Attack Defense on RAP
0Attack Success Rate (ASR)SENTINEL
Evaluation Results
| Method | Links | |
|---|---|---|
| SENTINELModel=GPT-4o-mini2026.07 | 0 | |
| A-MemG. (heur.)Model=GPT-4o-mini2026.07 | 38 | |
| NoDef.Model=GPT-4o-mini2026.07 | 48 | |
| KeyFilt.Model=GPT-4o-mini2026.07 | 48 |
| Method | Links | |
|---|---|---|
| SENTINELModel=GPT-4o-mini2026.07 | 0 | |
| A-MemG. (heur.)Model=GPT-4o-mini2026.07 | 38 | |
| NoDef.Model=GPT-4o-mini2026.07 | 48 | |
| KeyFilt.Model=GPT-4o-mini2026.07 | 48 |