Simple Reasoning on CSQA
91.75AccuracyBaseline
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BaselineTopology=Star2025.11 | 91.75 | — | |
| BaselineTopology=Chain2025.11 | 91.46 | — | |
| BaselineTopology=Complete2025.11 | 90.93 | — | |
| MAS-ShieldTopology=Star2025.11 | 90.78 | 22.05 | |
| BaselineTopology=Tree2025.11 | 90.68 | — | |
| MAS-ShieldTopology=Chain2025.11 | 90.64 | 22.78 | |
| MAS-ShieldTopology=Complete2025.11 | 90.13 | 21.36 | |
| MAS-ShieldTopology=Tree2025.11 | 89.31 | 24.04 | |
| BaselineTopology=Circle2025.11 | 88.91 | — | |
| MAS-ShieldTopology=Circle2025.11 | 87.55 | 22.11 | |
| AttackTopology=Complete2025.11 | 68.79 | — | |
| AttackTopology=Star2025.11 | 68.73 | — | |
| AttackTopology=Chain2025.11 | 67.86 | — | |
| AttackTopology=Circle2025.11 | 65.44 | — | |
| AttackTopology=Tree2025.11 | 65.27 | — |