Tag
This paper evaluates GPT-4.1's ability to simulate personas for opinion prediction, achieving accurate election forecasts in eight out of nine U.S. states and high accuracy in predicting beliefs about childhood vaccines, though simulated dialogues lacked natural human flow.
This paper proposes AI Tour Meeting, a group travel planning framework that uses multiple LLM-based agents with distinct personas to collaboratively find itineraries through natural language discussion.
Researchers from KAIST propose a framework that uses persona-guided LLM agents to synthesize diverse harmful content for stress-testing detection systems, addressing limitations of static benchmarks such as scalability, diversity, and data contamination. Both human and LLM evaluations confirm the synthetic scenarios are harder to detect than existing benchmarks while maintaining linguistic and topical diversity.