Response Safety Assessment on TTI Attack Dataset Responses
49Safe Response Countanthropic-claude-3.5-haiku
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| anthropic-claude-3.5-haiku2026.04 | 49 | 1 | 98 | 2 | |
| openai-gpt-4.1-mini2026.04 | 46 | 4 | 92 | 8 | |
| openai-gpt-4o-2024-08-062026.04 | 46 | 4 | 92 | 8 | |
| openai-gpt-4o-mini2026.04 | 46 | 4 | 92 | 8 | |
| meta-llama-llama-4-maverick2026.04 | 45 | 5 | 90 | 10 | |
| deepseek-deepseek-r1-05282026.04 | 45 | 5 | 90 | 10 | |
| mistral-medium-32026.04 | 44 | 6 | 88 | 12 | |
| mistral-small-3.1-24b-instruct2026.04 | 40 | 10 | 80 | 20 | |
| mistral-devstral-small2026.04 | 38 | 12 | 76 | 24 | |
| gemini-2.0-flash2026.04 | 33 | 17 | 66 | 34 | |
| gemini-1.5-flash-8b2026.04 | 32 | 18 | 64 | 36 | |
| gemini-2.0-flash-lite2026.04 | 30 | 20 | 60 | 40 | |
| gemini-1.5-flash2026.04 | 30 | 20 | 60 | 40 |