financial-compliance

Introduces ReguSim and ReguBench to evaluate LLM agent rule grounding in financial compliance, finding that visible rules reduce but do not eliminate violations and that incentive framing affects behavior.

This paper proposes a cycle-consistent neural architecture that generates faithful natural language explanations of formal verification certificates, achieving 90% soundness and 860x faster inference than LLM baselines.

This article argues that the accountability infrastructure developed for automated financial trading over decades provides a blueprint for governing AI agents, drawing on specific regulatory mechanisms like FINRA Rule 5310 and SEC Rule 17a-4.