Memory Attack on CSQA
1.67ASR@3G-Safeguard
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| G-SafeguardTopology=Star, Backbone LLM=GPT-4o-mini2025.08 | 1.67 | 96.44 | |
| G-SafeguardTopology=Random, Backbone LLM=GPT-4o-mini2025.08 | 5 | 95.56 | |
| G-SafeguardTopology=Tree, Backbone LLM=GPT-4o-mini2025.08 | 6.67 | 93.78 | |
| BlindGuardTopology=Chain, Backbone LLM=GPT-4o-mini2025.08 | 7 | 73.78 | |
| G-SafeguardTopology=Chain, Backbone LLM=GPT-4o-mini2025.08 | 7.67 | 94.67 | |
| BlindGuardTopology=Tree, Backbone LLM=GPT-4o-mini2025.08 | 7.67 | 72.89 | |
| BlindGuardTopology=Random, Backbone LLM=GPT-4o-mini2025.08 | 8.67 | 75.56 | |
| BlindGuardTopology=Star, Backbone LLM=GPT-4o-mini2025.08 | 9.33 | 76.44 | |
| TAMTopology=Tree, Backbone LLM=GPT-4o-mini2025.08 | 13 | 54.67 | |
| TAMTopology=Chain, Backbone LLM=GPT-4o-mini2025.08 | 13.33 | 51.11 | |
| PREMTopology=Tree, Backbone LLM=GPT-4o-mini2025.08 | 14 | 64.44 | |
| PREMTopology=Chain, Backbone LLM=GPT-4o-mini2025.08 | 15.67 | 68.89 | |
| PREMTopology=Random, Backbone LLM=GPT-4o-mini2025.08 | 16.67 | 64.44 | |
| DOMINANTTopology=Chain, Backbone LLM=GPT-4o-mini2025.08 | 18.67 | 36.44 | |
| TAMTopology=Star, Backbone LLM=GPT-4o-mini2025.08 | 19 | 61.33 | |
| PREMTopology=Star, Backbone LLM=GPT-4o-mini2025.08 | 20 | 70.22 | |
| No DefenseTopology=Tree, Backbone LLM=GPT-4o-mini2025.08 | 21.33 | — | |
| DOMINANTTopology=Tree, Backbone LLM=GPT-4o-mini2025.08 | 21.67 | 40.44 | |
| No DefenseTopology=Random, Backbone LLM=GPT-4o-mini2025.08 | 24 | — | |
| No DefenseTopology=Chain, Backbone LLM=GPT-4o-mini2025.08 | 26.67 | — | |
| DOMINANTTopology=Star, Backbone LLM=GPT-4o-mini2025.08 | 26.67 | 36.44 | |
| TAMTopology=Random, Backbone LLM=GPT-4o-mini2025.08 | 26.67 | 52 | |
| No DefenseTopology=Star, Backbone LLM=GPT-4o-mini2025.08 | 27 | — | |
| DOMINANTTopology=Random, Backbone LLM=GPT-4o-mini2025.08 | 34.33 | 38.22 |