Mathematical Reasoning on GSM8K (Attack & Defense Accuracy)
100Accuracy under Attack (GSM8K)Reporting-and-penalty mechanism
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Reporting-and-penalty mechanismAttack Type=All Attack Types, LLM Model=All Above Models2026.04 | 100 | 100 | |
| SentinelNetAttack Type=Collaboration Attack, LLM Model=Qwen2.5-72B-128K2026.04 | 70 | 97 | |
| GroupGuardAttack Type=False Consensus, LLM Model=Deepseek-V3.12026.04 | 60 | 99 | |
| GroupGuardAttack Type=False Consensus, LLM Model=Qwen3-32B2026.04 | 57 | 100 | |
| GroupGuardAttack Type=False Consensus, LLM Model=Qwen3-235B2026.04 | 55 | 81 |