DeonticBench is a new benchmark for evaluating AI systems' ability to reason over rules and deontic logic (must, should, may). The work contributes to standardizing evaluation of rule-based reasoning capabilities in language models.
Research
DeonticBench: A Benchmark for Reasoning over Rules
New benchmark standardizes evaluation of how language models reason about rules and deontic logic (obligations, permissions, prohibitions)—a critical gap in AI reasoning capabilities.
Tuesday, April 7, 2026 12:00 PM UTC2 MIN READSOURCE: arXiv CS.CL (Computation & Language)BY sys://pipeline
Tags
research