| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| NQ (Natural Questions) | VectorSteer | ORR0 | 72 | 3mo ago | |
| ORFuzzSet | LLM-VA | ORR16 | 72 | 3mo ago | |
| Over-refusal (test) | No train | Refusal Rate0 | 36 | 1mo ago | |
| MMMU in-scope (test) | Prompt-based | Math Score37 | 32 | 3mo ago | |
| ScienceQA in-scope (test) | System Prompt | Biology Refusal Count0 | 32 | 3mo ago | |
| XSTest | Evaluation Score (avg@4)100 | 26 | 1d ago | ||
| XSTest Safe | AdaCD | Over-refusal Rate1.6 | 25 | 1mo ago | |
| OR-Bench (boundary cases) | Mistral-v2-7B | OR-FPR1.7 | 18 | 8d ago | |
| WildGuard Unharmful | Categorical Steering | Over-refusal Rate1.06 | 7 | 2mo ago | |
| CoCoNot Contrast | Categorical Steering | Over-refusal Rate1.58 | 7 | 2mo ago | |
| Benign prompt dataset | Base | Over-Refusal Rate17 | 4 | 1mo ago | |
| XSTest (test) | Claude Sonnet 4.5 | Over-refusal Rate0.035 | 4 | 3mo ago |