DAR: Deontic Reasoning with Agentic Harnesses
This work addresses the challenge of applying long, cross-referenced rules in deontic reasoning for LLMs, but improvements are incremental and inconsistent.
DAR introduces an agentic reasoning setup for deontic reasoning tasks, where LLMs interact with statutes on demand. On hard subsets of DeonticBench, agentic harnesses improve performance but not uniformly, with weaker models degrading on numerical tasks and consuming more tokens.
Deontic reasoning is the task of answering questions by applying explicit rules and policies to case-specific facts, for example computing tax liability under a statute or determining the outcome of an immigration appeal. A key technical challenge for LLM-based deontic reasoning is that the relevant ruleset can be long and cross-referenced, so models may still fail to locate the rules needed for a particular reasoning step. We introduce Deontic Agentic Reasoning (DAR), an agentic reasoning setup in which the model interacts with the statutes on demand. We evaluate DAR under multiple harnesses on hard subsets of DeonticBench. Across these settings, we find that agentic harnesses can push the frontier on deontic reasoning tasks, but improvements are not uniform: weaker models often degrade on numerical tasks while consuming far more tokens.