Policy Scope Adaptation on Adaptive Policy Scope Benchmark Expanding-Narrowing Categories
96RecallQwen3-8B-Thinking
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-8B-ThinkingParameter Scale=8B, Reasoning Mode=Thinking2026.01 | 96 | 56 | 71 | |
| Qwen3-32B-ThinkingParameter Scale=32B, Reasoning Mode=Thinking2026.01 | 96 | 64 | 77 | |
| GPT-OSS-SafeGuard-20BParameter Scale=20B2026.01 | 95 | 52 | 67 | |
| Qwen3-32B-NoThinkingParameter Scale=32B, Reasoning Mode=NoThinking2026.01 | 95 | 50 | 66 | |
| YuFeng-XGuardReasoning Mode=Implicit2026.01 | 89 | 65 | 75 | |
| Qwen3-8B-NoThinkingParameter Scale=8B, Reasoning Mode=NoThinking2026.01 | 4 | 50 | 7 |