arXiv cs.AI· Yiyang Luo, Yihang Jiang, Qijun Xie, Liang Lan, Lin Willian Cong, Anyi Rao, Yunya Song·· 3 天前AI 评分31
ReguSim：评估 LLM 智能体在金融合规中的规则落地能力
ReguSim: Evaluating LLM Agent Rule Grounding in Financial Compliance
AI 导读
研究者推出 ReguSim 金融合规环境和 ReguBench 目标标记监测基准，用于区分陈述推理、尝试动作、执行约束与监测证据四个环节。在 DeepSeek V4 Pro 和 Gemini 3.5 Flash 的交易智能体测试中，可见规则能减少但无法消除被拒动作，激励或人设设定会改变行为。监测环节中，简单结构化基线可持平或超过仅用提示词的 LLM。
正文 · 原文
Abstract:LLM agents in financial markets may cite rules yet still submit orders that violate executable constraints or misread surveillance evidence. We introduce ReguSim, a controlled financial-compliance environment, and ReguBench, a target-marked monitoring benchmark, to separate four artifacts: stated reasoning, attempted action, execution enforcement, and monitor evidence. In trader runs with DeepSeek V4 Pro and Gemini 3.5 Flash, visible rules reduce but do not eliminate rejected actions, and incentive or persona framing shifts behavior. A bridge study shows that trader rationales can mislead an independent monitor unless enforcement evidence is shown. In monitoring, simple structured baselines either match or exceed prompt-only LLMs. The results frame financial compliance evaluation as an audit of rule-grounded actions and evidence use, rather than a single compliance score.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.19974 [cs.AI] |
| (or arXiv:2608.19974v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2608.19974 arXiv-issued DOI via DataCite |
Submission history
From: Yiyang Luo [view email]
[v1]
Thu, 20 Aug 2026 12:49:05 UTC (382 KB)
[v2]
Mon, 12 Oct 2026 07:03:25 UTC (375 KB)
来源：arXiv cs.AI · arxiv.org