Discriminative Evaluation on RuDE base
93.1ADDeepSeek-V3.1
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DeepSeek-V3.1Violation Number (|V|)=3 for WR, 5 for others2026.05 | 93.1 | 84.3 | 73.1 | 64.6 | 78.8 | |
| Seed-OSS-36BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=36B2026.05 | 86.6 | 76.8 | 63 | 58.4 | 71.2 | |
| Qwen3-235B-A22BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=235B-A22B2026.05 | 85.8 | 72.7 | 68.5 | 56.5 | 70.9 | |
| GLM-4.5Violation Number (|V|)=3 for WR, 5 for others2026.05 | 85.7 | 78.1 | 65 | 72 | 75.2 | |
| Kimi-K2Violation Number (|V|)=3 for WR, 5 for others2026.05 | 83.5 | 77.7 | 80.1 | 57.6 | 74.7 | |
| Qwen2.5-72BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=72B2026.05 | 80.5 | 67.1 | 57.4 | 51.8 | 64.2 | |
| Qwen3-14BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=14B2026.05 | 77.7 | 62.9 | 62.9 | 51.4 | 63.7 | |
| Qwen3-8BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=8B2026.05 | 75.6 | 60.6 | 60 | 49.4 | 61.4 | |
| Gemma3-27BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=27B2026.05 | 74.9 | 65.2 | 57.3 | 49.9 | 61.8 | |
| Qwen3-30B-A3BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=30B-A3B2026.05 | 74.5 | 62.8 | 59 | 49.8 | 61.5 | |
| GLM-4.5-AirViolation Number (|V|)=3 for WR, 5 for others2026.05 | 73.4 | 66.4 | 57.9 | 50.1 | 62 | |
| Gemma3-12BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=12B2026.05 | 71.3 | 57.4 | 55.3 | 48.2 | 58.1 | |
| Qwen2.5-14BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=14B2026.05 | 69.9 | 63.6 | 59.9 | 50 | 60.9 | |
| Qwen3-4BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=4B2026.05 | 64.4 | 53.9 | 50.1 | 45.7 | 53.5 | |
| Gemma3-4BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=4B2026.05 | 51 | 51.4 | 41.3 | 37.9 | 45.4 | |
| Qwen2.5-7BViolation Number (|V|)=3 for WR, 5 for others, Parameter Count=7B2026.05 | 43.4 | 40.7 | 42.1 | 39.5 | 41.4 |