Qualitative Safety Assessment on KT AI Risk Taxonomy prompts
99.38Not Unsafe Rate (Content Safety)HyperCLOVAX-SEED-Think-32B
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| HyperCLOVAX-SEED-Think-32BReasoning=Off2026.03 | 99.38 | 98.25 | 94.17 | 97.55 | 78.79 | 63.64 | 92.86 | 76.81 | |
| Mi:dm K 2.5 ProReasoning=Off2026.03 | 98.75 | 95 | 94.17 | 96.14 | 100 | 96.45 | 100 | 98.55 | |
| Mi:dm K 2.5 ProReasoning=On2026.03 | 98.38 | 94.63 | 95.67 | 96.27 | 100 | 100 | 100 | 100 | |
| Qwen-3-30B-A3BReasoning=On2026.03 | 96.75 | 93.5 | 92.17 | 94.32 | 100 | 100 | 100 | 100 | |
| HyperCLOVAX-SEED-Think-32BReasoning=On2026.03 | 96.75 | 94.63 | 91 | 94.41 | 100 | 100 | 100 | 100 | |
| EXAONE-4.0-32BReasoning=On2026.03 | 93 | 90.25 | 84.83 | 89.77 | 100 | 100 | 100 | 100 | |
| Qwen-3-30B-A3BReasoning=Off2026.03 | 90.88 | 91 | 78.83 | 87.64 | 100 | 95.45 | 100 | 98.55 | |
| EXAONE-4.0-32BReasoning=Off2026.03 | 82.88 | 83 | 75.5 | 79.55 | 100 | 95.45 | 100 | 98.55 |