用分工协作的智能体团队实现安全开放探索，失败经验还能变成可复用的约束。
Heterogeneous Agent Cohorts for Safe Open-Ended Exploration with Runtime Constraint Memory
- 拆分角色：破坏者出新点子，验证者守安全底线，中介者引外部类比。
- 实验中零执行违规，节省15.1%令牌，降低55.9%通信开销。
- 失败记录生成可继承的‘伤疤’约束，适合资源受限的安全强化学习场景。
当前大模型智能体面临两难：严格限制则缺乏创意，自由探索又易引发安全问题。本文提出将职责分离：由破坏者生成非常规方案，验证者在工具入口执行实时检查，中介者引入远距离相关类比。失败不被丢弃，而是通过蒙特卡洛树搜索（MCTS）提炼为紧凑的、带签名的约束补丁，称为“伤疤”（Scars）。这些补丁本地缓存并传递给后续智能体组，使重复失败转化为低成本的运行时约束。在空间语义沙盒中（N=20次运行，p<0.01），该方法成功抵达辩论机制无法到达的目标，验证者阻止所有执行违规，且伤疤使令牌消耗减少15.1%。此外，基于信用的通信分配评分（CAS）限制输出带宽，在资源受限下整体令牌成本降低55.9%。
原文摘要 · Abstract (English)
LLM agents today are caught in an awkward bind. Lock them down with static safety instructions and they rarely venture beyond the obvious; give them free reign with tools and multi-agent debate, and safety violations quickly follow. Rather than forcing a single model to juggle both creativity and caution, we separate the concerns across specialized roles. A Disrupter generates unconventional proposals, a Validator enforces hard runtime checks at the tool gateway, and a Broker pulls in distant but relevant analogies. Failures are not discarded -- they are compiled, via MCTS, into compact, signed constraint patches we call Scars. These patches are cached locally and inherited by future cohorts, turning repeated failures into reusable, low-cost runtime constraints. In a spatial-semantic sandbox (N=20 runs, p<0.01), our cohort reaches remote targets where debate fails, the Validator prevents all executed breaches, and Scars reduce token consumption by 15.1% by avoiding redundant validator checks. Furthermore, credit-based Communication Allocation Scores (CAS) restrict outbound bandwidth, reducing overall token costs by 55.9% under resource constraints.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。