Prompt Safety Assessment on TTI Attack Dataset Prompts
30Safe Prompt Countgemini-2.0-flash
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| gemini-2.0-flash2026.04 | 30 | 20 | 60 | 40 | |
| gemini-2.0-flash-lite2026.04 | 30 | 20 | 60 | 40 | |
| gemini-1.5-flash2026.04 | 30 | 20 | 60 | 40 | |
| gemini-1.5-flash-8b2026.04 | 29 | 21 | 58 | 42 | |
| anthropic-claude-3.5-haiku2026.04 | 22 | 28 | 44 | 56 | |
| mistral-medium-32026.04 | 21 | 29 | 42 | 58 | |
| openai-gpt-4.1-mini2026.04 | 21 | 29 | 42 | 58 | |
| openai-gpt-4o-mini2026.04 | 21 | 29 | 42 | 58 | |
| mistral-devstral-small2026.04 | 20 | 30 | 40 | 60 | |
| mistral-small-3.1-24b-instruct2026.04 | 20 | 30 | 40 | 60 | |
| openai-gpt-4o-2024-08-062026.04 | 20 | 30 | 40 | 60 | |
| deepseek-deepseek-r1-05282026.04 | 20 | 30 | 40 | 60 | |
| meta-llama-llama-4-maverick2026.04 | 18 | 32 | 36 | 64 |