URL Vulnerability Evaluation on MalURLBench 1.0 (test)
100Success Rate (Commercial)Mixtral-8x7b
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Mixtral-8x7bOptimization=false2026.01 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | |
| Llama2-7B-chat-hfOptimization=false2026.01 | 100 | 94 | 98 | 100 | 94 | 99 | 69 | 96 | 99 | 100 | |
| Mistral-smallOptimization=false2026.01 | 98 | 94 | 100 | 100 | 98 | 97 | 100 | 95 | 100 | 100 | |
| GPT-4o-miniOptimization=false2026.01 | 97 | 66 | 100 | 100 | 100 | 91 | 87 | 64 | 100 | 100 | |
| Deepseek-chatOptimization=false2026.01 | 92 | 78 | 100 | 75 | 50 | 81 | 90 | 88 | 97 | 94 | |
| Llama-3-8bOptimization=false2026.01 | 72 | 58 | 53 | 100 | 10 | 32 | 33 | 7 | 10 | 95 | |
| Llama-3-8bOptimization=true2026.01 | 72 | 58 | 53 | 100 | 20 | 32 | 33 | 33 | 10 | 95 | |
| Mistral-7bOptimization=false2026.01 | 70 | 46 | 63 | 60 | 38 | 16 | 52 | 48 | 53 | 65 | |
| Deepseek-coderOptimization=false2026.01 | 33 | 6 | 1 | 72 | 57 | 17 | 8 | 15 | 46 | 85 | |
| Deepseek-coderOptimization=true2026.01 | 33 | 33 | 27 | 72 | 57 | 17 | 32 | 15 | 46 | 85 | |
| GPT-3.5-turboOptimization=true2026.01 | 23 | 39 | 40 | 41 | 20 | 13 | 27 | 34 | 42 | 50 | |
| Llama-3-70bOptimization=true2026.01 | 21 | 41 | 34 | 66 | 26 | 34 | 26 | 47 | 21 | 41 | |
| GPT-4oOptimization=false2026.01 | 18 | 13 | 5 | 74 | 69 | 29 | 37 | 1 | 22 | 77 | |
| GPT-4oOptimization=true2026.01 | 18 | 13 | 12 | 74 | 69 | 29 | 37 | 31 | 22 | 77 | |
| URLGuardOptimization=false2026.01 | 14 | 5 | 1 | 61 | 0 | 0 | — | 0 | 0 | 70 | |
| Qwen-plusOptimization=true2026.01 | 12 | 22 | 23 | 51 | 25 | 68 | 12 | 20 | 51 | 73 | |
| Llama-3-70bOptimization=false2026.01 | 3 | 2 | 34 | 66 | 26 | 5 | 26 | 7 | 21 | 41 | |
| GPT-3.5-turboOptimization=false2026.01 | 1 | 6 | 4 | 41 | 9 | 13 | 9 | 0 | 8 | 50 | |
| Qwen-plusOptimization=false2026.01 | 1 | 0 | 23 | 51 | 1 | 68 | 12 | 20 | 51 | 73 |