Attack Detection on Leave-one-model-out held-out 9th model 1.0 (val)
1AUCRandom Forest
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Random ForestHeld-Out Model=glm-4.7-flash:q8.02026.06 | 1 | 100 | |
| Random ForestHeld-Out Model=gpt-oss-safeguard:120b2026.06 | 1 | 100 | |
| Random ForestHeld-Out Model=gpt-oss:20b2026.06 | 1 | 100 | |
| Random ForestHeld-Out Model=qwen2.5:72b2026.06 | 1 | 100 | |
| Random ForestHeld-Out Model=qwen3.5:9b2026.06 | 1 | 100 | |
| Random ForestHeld-Out Model=qwen3:32b2026.06 | 1 | 100 | |
| Random ForestHeld-Out Model=qwen3.5:122b2026.06 | 0.75 | 100 | |
| Random ForestHeld-Out Model=qwen2.5:14b2026.06 | 0.083 | 100 | |
| Random ForestHeld-Out Model=qwq:32b2026.06 | 0 | 0 |