Output-refinement on Tweet Engagement
85TGRVanilla
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VanillaModel=LLaMA-3.1-405B2026.04 | 85 | 3.96 | 0.0004 | |
| VanillaModel=GPT-42026.04 | 76 | 5.15 | 2.24 | |
| Self DefenseModel=GPT-42026.04 | 74.46 | 5.56 | 6.44 | |
| Goal PriorityModel=GPT-42026.04 | 70 | 3.69 | 0.0007 | |
| Goal PriorityModel=LLaMA-3.1-405B2026.04 | 70 | 1.69 | 0.0535 | |
| VanillaModel=Qwen2.5-72B2026.04 | 68.75 | 3.51 | 0.0007 | |
| Self DefenseModel=Qwen2.5-72B2026.04 | 60 | 1.25 | 0.113 | |
| Self DefenseModel=GPT-3.52026.04 | 58.76 | 1.21 | 0.11 | |
| Self DefenseModel=LLaMA-3.1-405B2026.04 | 57.89 | 0.64 | 0.263 | |
| VanillaModel=GPT-3.52026.04 | 54.54 | 2.63 | 0.004 | |
| LCOModel=LLaMA-3.1-405B2026.04 | 52.63 | 0.06 | 0.477 | |
| LCOModel=GPT-3.52026.04 | 46.15 | -0.55 | 0.7 | |
| LCOModel=Qwen2.5-72B2026.04 | 42.85 | 0.33 | 0.374 | |
| Goal PriorityModel=GPT-3.52026.04 | 42.1 | -0.11 | 0.543 | |
| LCOModel=GPT-42026.04 | 36.73 | -2.36 | 0.99 | |
| Goal PriorityModel=Qwen2.5-72B2026.04 | 35 | -0.74 | 0.767 |