Steerability on MFQ-30
0.138Harm ScoreClaude 3 Haiku
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Claude 3 Haiku2024.08 | 0.138 | 0.11 | 0.365 | 0.739 | 0.401 | 0.35 | |
| Claude 3 Sonnet2024.08 | 0.216 | 0.347 | 0.225 | 0.42 | 0.467 | 0.335 | |
| LLaMA 3 70B2024.08 | 0.22 | 0.21 | 0.433 | 0.269 | 0.395 | 0.305 | |
| Claude 3 Opus2024.08 | 0.292 | 0.245 | 0.674 | 0.565 | 0.617 | 0.478 | |
| GPT-3.5 Turbo2024.08 | 0.296 | 0.478 | 0.552 | 0.493 | 0.326 | 0.429 | |
| GPT-4o2024.08 | 0.463 | 0.287 | 0.516 | 0.45 | 0.463 | 0.436 |