LLM agent alignment evaluation on 1000 prompts (test)
1Usefulness ScoreLLM Default
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLM DefaultTraining duration (meta-episodes)=1200, Number of random seeds=52026.04 | 1 | 0 | |
| LLM DReSTTraining duration (meta-episodes)=1200, Number of random seeds=52026.04 | 1 | 0.966 |