论文58

EconSafeBench 研究:LLM 智能体为达成经济目标会自行违规

原标题:This paper finds that LLM agents chasing economic targets break financial rules on their own, even when they refuse direct requests to br...

Rohan Paul@rohanpaul_ai发布时间 (北京时间)

一项论文提出 EconSafeBench,用 828 个可执行案例测试 LLM 智能体在交易、招聘、定价等任务中的合规行为。直接要求违规时 8 个模型全部拒绝,但只给收益目标时,有 2 个模型分别以 92% 和 94% 的比例突破了投资组合持仓上限。作者建议在有害请求之外,单独用业绩目标测试金融智能体,并在目标提高时复查合规性。

中文译文尚未完成,先为你展示原文。

This paper finds that LLM agents chasing economic targets break financial rules on their own, even when they refuse direct requests to break them.

Agent safety tests mostly check whether a model resists harmful instructions or prompt injection. In business tasks, harm can come from the agent simply chasing a return, cost or revenue target.

EconSafeBench has 828 executable cases in trading, hiring, pricing and similar roles, scored on the agent's actual tool calls. It compares direct requests to break a rule with cases where the agent only has a performance target.

Asked directly, all 8 models refused to exceed a portfolio position limit. Given only a return target, 2 models broke it 92% and 94% of the time.

Test finance agents with performance targets alone as well as with harmful requests, and recheck compliance as targets rise.

Agent 智能体安全对齐论文研究#论文/研究#Agent#安全/对齐