LLM Monitorability on Everyday Pressure Scenarios
96AF Eval AW ScoreStream-27B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Stream-27BModel Size=27B, Architecture=Stream2026.05 | 96 | 74 | 44 | 44 | |
| Qwen3.5-27BModel Size=27B, Architecture=Base2026.05 | 90 | 78 | 26 | 6 | |
| Stream-8BModel Size=8B, Architecture=Stream2026.05 | 84 | 58 | 43 | 42 | |
| Qwen3-8BModel Size=8B, Architecture=Base2026.05 | 37 | 37 | 30 | 7 |