Evaluation Awareness Detection on Apollo Evaluation Traces (Scheming samples)
8.9Detection Rate (Scheming Samples)gpt-5-thinking
Evaluation Results
| Method | Links | |
|---|---|---|
| gpt-5-thinkingEvaluation awareness definition=Broad2025.12 | 8.9 | |
| gpt-5-thinkingEvaluation awareness definition=Strict2025.12 | 5 |