Researchllmsocial deductionevaluation
Carrot-Parsnip Introduces Social Deduction LLM Evals
5.8
Relevance Score
LessWrong presents Carrot-Parsnip, a social deduction game designed to evaluate large language models' reasoning about hidden player roles; SD games require players to reason about other players' concealed roles and group dynamics.
Scoring Rationale
Moderate novelty and relevance, but RSS-only limited details and single-source origin reduce confidence and practical impact estimation.
Sources
- Read OriginalCarrot-Parsnip: A Social Deduction Game for LLM Evals — LessWronglesswrong.com
