Researchllmsocial deductionevaluation
Carrot-Parsnip Introduces Social Deduction LLM Evals
5.8
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.
Key Points
- 1Introduces Carrot-Parsnip, a social deduction game tailored for evaluating LLM behavior and reasoning.
- 2Explains SD games require reasoning about other players' hidden roles and ambiguous information.
- 3Suggests a new evaluation angle probing models' inference under partial-information and limited context.
Scoring Rationale
Moderate novelty and relevance, but RSS-only limited details and single-source origin reduce confidence and practical impact estimation.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
Used by DS/ML engineers at top companies
High-Value Overnight OrdersEasyDelivered International ShipmentsMediumOn-Time Delivery Rate by CarrierHard
250 free problems · No credit card
See all Logistics & Shipping problems
