Physics Question Answering on MMLU Phy
94.6ASR (Accuracy)No trust
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| No trustCommunication Structure=Complete, Attack Type=AiTM2025.06 | 94.6 | — | |
| No trustCommunication Structure=Chain, Attack Type=AiTM2025.06 | 92.5 | — | |
| No trustCommunication Structure=Tree, Attack Type=AiTM2025.06 | 87.6 | — | |
| No trustCommunication Structure=Complete, Attack Type=AutoInj2025.06 | 74.5 | — | |
| No trustCommunication Structure=Chain, Attack Type=AutoTrans2025.06 | 69.3 | — | |
| No trustCommunication Structure=Complete, Attack Type=AutoTrans2025.06 | 68.3 | — | |
| No trustCommunication Structure=Chain, Attack Type=NetSafe2025.06 | 67.7 | — | |
| PPLCommunication Structure=Complete, Attack Type=AiTM2025.06 | 67.2 | — | |
| No trustCommunication Structure=Complete, Attack Type=NetSafe2025.06 | 63.9 | — | |
| PromptCommunication Structure=Complete, Attack Type=AiTM2025.06 | 63.7 | — | |
| PPLCommunication Structure=Tree, Attack Type=AiTM2025.06 | 63.7 | — | |
| PromptCommunication Structure=Tree, Attack Type=AiTM2025.06 | 62.9 | — | |
| No trustCommunication Structure=Tree, Attack Type=AutoInj2025.06 | 62.8 | — | |
| No trustCommunication Structure=Chain, Attack Type=AutoInj2025.06 | 62.6 | — | |
| PPLCommunication Structure=Chain, Attack Type=AutoTrans2025.06 | 62.6 | — | |
| PromptCommunication Structure=Chain, Attack Type=AutoTrans2025.06 | 60.8 | — | |
| PPLCommunication Structure=Complete, Attack Type=AutoTrans2025.06 | 59.5 | — | |
| PPLCommunication Structure=Chain, Attack Type=AutoInj2025.06 | 59.4 | — | |
| PromptCommunication Structure=Chain, Attack Type=NetSafe2025.06 | 58.7 | — | |
| PPLCommunication Structure=Complete, Attack Type=AutoInj2025.06 | 58.4 | — | |
| No trustCommunication Structure=Tree, Attack Type=AutoTrans2025.06 | 58.4 | — | |
| PPLCommunication Structure=Chain, Attack Type=NetSafe2025.06 | 58.3 | — | |
| No trustCommunication Structure=Tree, Attack Type=NetSafe2025.06 | 57.8 | — | |
| PromptCommunication Structure=Complete, Attack Type=AutoInj2025.06 | 57.7 | — | |
| PromptCommunication Structure=Chain, Attack Type=AutoInj2025.06 | 57.3 | — | |
| PromptCommunication Structure=Tree, Attack Type=AutoInj2025.06 | 57.3 | — | |
| PromptCommunication Structure=Complete, Attack Type=AutoTrans2025.06 | 56.8 | — | |
| PPLCommunication Structure=Complete, Attack Type=NetSafe2025.06 | 56.6 | — | |
| PPLCommunication Structure=Tree, Attack Type=AutoInj2025.06 | 56.6 | — | |
| PPLCommunication Structure=Tree, Attack Type=AutoTrans2025.06 | 53.8 | — | |
| PromptCommunication Structure=Complete, Attack Type=NetSafe2025.06 | 53.7 | — | |
| PromptCommunication Structure=Tree, Attack Type=AutoTrans2025.06 | 52.7 | — | |
| PPLCommunication Structure=Tree, Attack Type=NetSafe2025.06 | 52.2 | — | |
| PromptCommunication Structure=Tree, Attack Type=NetSafe2025.06 | 51.8 | — | |
| A-TrustCommunication Structure=Chain, Attack Type=AutoTrans2025.06 | 51.4 | — | |
| A-TrustCommunication Structure=Complete, Attack Type=AutoInj2025.06 | 49.1 | — | |
| A-TrustCommunication Structure=Complete, Attack Type=AutoTrans2025.06 | 47.9 | — | |
| PPLCommunication Structure=Chain, Attack Type=AiTM2025.06 | 47.3 | — | |
| A-TrustCommunication Structure=Chain, Attack Type=AutoInj2025.06 | 47.1 | — | |
| A-TrustCommunication Structure=Complete, Attack Type=NetSafe2025.06 | 46.8 | — | |
| A-TrustCommunication Structure=Chain, Attack Type=NetSafe2025.06 | 46.4 | — | |
| A-TrustCommunication Structure=Tree, Attack Type=AutoInj2025.06 | 46.4 | — | |
| A-TrustCommunication Structure=Tree, Attack Type=NetSafe2025.06 | 46.3 | — | |
| A-TrustCommunication Structure=Tree, Attack Type=AutoTrans2025.06 | 45.5 | — | |
| PromptCommunication Structure=Chain, Attack Type=AiTM2025.06 | 38.2 | — | |
| A-TrustCommunication Structure=Tree, Attack Type=AiTM2025.06 | 24.8 | — | |
| A-TrustCommunication Structure=Complete, Attack Type=AiTM2025.06 | 23.5 | — | |
| A-TrustCommunication Structure=Chain, Attack Type=AiTM2025.06 | 14.1 | — | |
| A-TrustCommunication Structure=Chain, Attack Type=Clean2025.06 | — | 43.8 | |
| A-TrustCommunication Structure=Complete, Attack Type=Clean2025.06 | — | 42.6 | |
| A-TrustCommunication Structure=Tree, Attack Type=Clean2025.06 | — | 42.3 | |
| No trustCommunication Structure=Chain, Attack Type=Clean2025.06 | — | 41.7 | |
| No trustCommunication Structure=Complete, Attack Type=Clean2025.06 | — | 39.6 | |
| No trustCommunication Structure=Tree, Attack Type=Clean2025.06 | — | 40.9 | |
| PPLCommunication Structure=Chain, Attack Type=Clean2025.06 | — | 41.9 | |
| PPLCommunication Structure=Complete, Attack Type=Clean2025.06 | — | 41.3 | |
| PPLCommunication Structure=Tree, Attack Type=Clean2025.06 | — | 40.6 | |
| PromptCommunication Structure=Chain, Attack Type=Clean2025.06 | — | 44.3 | |
| PromptCommunication Structure=Complete, Attack Type=Clean2025.06 | — | 41 | |
| PromptCommunication Structure=Tree, Attack Type=Clean2025.06 | — | 44.7 |