Vulnerability Prediction on Debate
0.81Spearman RhoMESA
Evaluation Results
| Method | Links | |
|---|---|---|
| MESAModel=Llama-8B, MESA Variant=Static2026.06 | 0.81 | |
| MESAModel=Llama-8B, MESA Variant=Combined2026.06 | 0.77 | |
| MESAModel=Gemma-26B, MESA Variant=Combined2026.06 | 0.5 | |
| MESAModel=Gemma-26B, MESA Variant=Static2026.06 | 0.47 | |
| MESAModel=Gemma-E4B, MESA Variant=Combined2026.06 | 0.13 | |
| MESAModel=Qwen-9B, MESA Variant=Static2026.06 | 0.09 | |
| MESAModel=Gemma-E4B, MESA Variant=Static2026.06 | 0.09 | |
| MESAModel=Qwen-9B, MESA Variant=Combined2026.06 | 0.03 |