Model Inertia measurement on MFQ-30
71.9HarmLLaMA 3 70B
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| LLaMA 3 70BSeeds=32024.08 | 71.9 | 80.1 | 33.1 | 22.1 | 26.1 | 46.6 | |
| GPT-4oSeeds=32024.08 | 58.9 | 56.6 | 23.1 | 26.9 | 16.3 | 36.4 | |
| Claude 3 HaikuSeeds=32024.08 | 51.6 | 56 | 18.6 | 15.6 | 12.1 | 30.8 | |
| Claude 3 SonnetSeeds=32024.08 | 48.5 | 51.5 | 28.9 | 26.3 | 20.8 | 35.2 | |
| GPT-3.5 TurboSeeds=32024.08 | 35.5 | 29.3 | 9.3 | 8 | 14.9 | 19.4 | |
| Claude 3 OpusSeeds=32024.08 | 29.9 | 43.3 | 16.6 | 22.7 | 21.5 | 26.8 | |
| LLaMA 3 8BSeeds=32024.08 | 26 | 32.7 | 11.3 | 8.5 | 4.3 | 16.5 |