SHARP: Unlocking Interactive Hallucination via Stance Transfer in Role-Playing LLMs
This work addresses a fundamental limitation in role-playing LLMs for social interaction scenarios, though it appears incremental as it builds on existing hallucination research.
The paper tackles the problem of interactive hallucination in role-playing LLMs by defining it through stance transfer and constructing the SHARP benchmark to simulate multi-role interactions, with experiments confirming the paradigm's effectiveness and challenging conventional mitigation solutions.
The advanced role-playing capabilities of Large Language Models (LLMs) have enabled rich interactive scenarios, yet existing research in social interactions neglects hallucination while struggling with poor generalizability and implicit character fidelity judgments. To bridge this gap, motivated by human behaviour, we introduce a generalizable and explicit paradigm for uncovering interactive patterns of LLMs across diverse worldviews. Specifically, we first define interactive hallucination through stance transfer, then construct SHARP, a benchmark built by extracting relations from commonsense knowledge graphs and utilizing LLMs' inherent hallucination properties to simulate multi-role interactions. Extensive experiments confirm our paradigm's effectiveness and stability, examine the factors that influence these metrics, and challenge conventional hallucination mitigation solutions. More broadly, our work reveals a fundamental limitation in popular post-training methods for role-playing LLMs: the tendency to obscure knowledge beneath style, resulting in monotonous yet human-like behaviors - interactive hallucination.