Task Routing on Security
0.01Cost ($)Force Weak
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Force WeakModel=qwen2026.01 | 0.01 | 63.5 | 94.1 | |
| Force WeakModel=openai2026.01 | 0.0207 | 95.8 | 92.9 | |
| CASTERModel=qwen2026.01 | 0.0257 | 6.2 | 95.2 | |
| Force StrongModel=qwen2026.01 | 0.0274 | — | 95.9 | |
| Force StrongModel=deepseek2026.01 | 0.0983 | — | 91.1 | |
| CASTERModel=deepseek2026.01 | 0.1105 | -12.4 | 94.8 | |
| Force WeakModel=deepseek2026.01 | 0.1118 | -13.7 | 89.3 | |
| Force WeakModel=gemini2026.01 | 0.1408 | 85.7 | 94.8 | |
| CASTERModel=openai2026.01 | 0.2234 | 54.3 | 93.9 | |
| CASTERModel=gemini2026.01 | 0.3064 | 68.9 | 96.2 | |
| Force WeakModel=claude2026.01 | 0.3346 | 84.8 | 94.4 | |
| Force StrongModel=openai2026.01 | 0.4886 | — | 93.7 | |
| CASTERModel=claude2026.01 | 0.9129 | 58.4 | 95.1 | |
| Force StrongModel=gemini2026.01 | 0.9837 | — | 95.6 | |
| Force StrongModel=claude2026.01 | 2.1948 | — | 94.3 |