Tag
The paper formulates persona diversification as a set-level conditioning problem to mitigate homogeneity in LLM outputs, evaluating methods that enhance creativity and diversity across tasks like the Alternative Uses Task.