📄 中文摘要
Mozi 是一种双层架构，旨在解决药物发现领域中工具使用治理不当和长时间可靠性差的问题。在依赖性强的制药流程中，自治代理常常会偏离可重复的轨迹，早期的幻觉会在后续阶段成倍增加失败的风险。通过建立一个受控的监督-工作者层级，Mozi 在灵活的生成性人工智能与计算生物学的确定性严谨性之间架起了桥梁。控制平面（Layer A）确保了基于角色的工具隔离，从而有效地管理工具的使用，提升了药物发现过程的可靠性和可控性。
Mozi：药物发现 LLM 代理的治理自主性
出处: Mozi: Governed Autonomy for Drug Discovery LLM Agents
发布: 2026年3月5日
📄 English Summary
Mozi: Governed Autonomy for Drug Discovery LLM Agents
Mozi presents a dual-layer architecture designed to address critical issues in drug discovery, specifically unconstrained tool-use governance and poor long-horizon reliability. In dependency-heavy pharmaceutical pipelines, autonomous agents often deviate into irreproducible trajectories, where early-stage hallucinations can exponentially compound into downstream failures. By establishing a governed supervisor-worker hierarchy, Mozi bridges the flexibility of generative AI with the deterministic rigor of computational biology. The Control Plane (Layer A) enforces role-based tool isolation, effectively managing tool usage and enhancing the reliability and controllability of the drug discovery process.