Safety Evaluation on SORRY-Bench (compliance rate)
26.9Compliance RateQwen3.5-35B-A3B
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3.5-35B-A3BJudge Version=default, Evaluation Mode=mut-avg2026.06 | 26.9 | |
| Qwen3.5-35B-A3BJudge Version=v0, Evaluation Mode=mut-avg2026.06 | 26.6 | |
| Qwen3.5-35B-A3BJudge Version=default, Evaluation Mode=base2026.06 | 17.9 | |
| Qwen3.5-35B-A3BJudge Version=v1, Evaluation Mode=mut-avg2026.06 | 17.7 | |
| Qwen3.5-35B-A3BJudge Version=v0, Evaluation Mode=base2026.06 | 16.4 | |
| Qwen3.5-35B-A3BJudge Version=v1, Evaluation Mode=base2026.06 | 13.9 | |
| Qwen3.5-35B-A3B-psysafeJudge Version=default, Evaluation Mode=mut-avg2026.06 | 0.2 | |
| Qwen3.5-35B-A3B-psysafeJudge Version=v0, Evaluation Mode=mut-avg2026.06 | 0.1 | |
| Qwen3.5-35B-A3B-psysafeJudge Version=default, Evaluation Mode=base2026.06 | 0 | |
| Qwen3.5-35B-A3B-psysafeJudge Version=v0, Evaluation Mode=base2026.06 | 0 | |
| Qwen3.5-35B-A3B-psysafeJudge Version=v1, Evaluation Mode=base2026.06 | 0 | |
| Qwen3.5-35B-A3B-psysafeJudge Version=v1, Evaluation Mode=mut-avg2026.06 | 0 |