Defender Robustness on English scam dialogues (END) (test)
16.5Success Rateqwen-max-latest
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| qwen-max-latestTOTAL=12002026.01 | 16.5 | 22.6 | 40.7 | 20.3 | |
| glm-4.6-thinkingTOTAL=12102026.01 | 8.3 | 22.7 | 52.9 | 16 | |
| llama-3.1-405b-instructTOTAL=12092026.01 | 8 | 20.3 | 58.8 | 12.9 | |
| grok-4TOTAL=12202026.01 | 4.2 | 43.7 | 35.6 | 16.6 | |
| ERNIE-4.0-8KTOTAL=12002026.01 | 2.3 | 35.7 | 29.7 | 32.3 | |
| SparkDesk-v3.5TOTAL=12092026.01 | 1.8 | 24.7 | 47 | 26.5 | |
| claude-sonnet-4-5-thinkingTOTAL=12192026.01 | 1.5 | 61.3 | 22.6 | 14.6 | |
| deepseek-r1TOTAL=12162026.01 | 0.8 | 30 | 43 | 26.2 |