Conversational Instruction Following on UltraChat
9.1Overall ScoreEPI
Evaluation Results
| Method | Links | |
|---|---|---|
| EPIBase Model=Gemma-2-9B, Number of runs=32026.04 | 9.1 | |
| Static IsolationBase Model=Gemma-2-9B, Number of runs=32026.04 | 8.8 | |
| EPIBase Model=LLaMA-3-8B, Number of runs=32026.04 | 8.6 | |
| Full SFTBase Model=Gemma-2-9B, Number of runs=32026.04 | 8.5 | |
| Static IsolationBase Model=LLaMA-3-8B, Number of runs=32026.04 | 8.3 | |
| Full SFTBase Model=LLaMA-3-8B, Number of runs=32026.04 | 8.1 |