COVENANT：用于对齐代理执行的自然语言工作流编译
COVENANT: Natural-Language Workflow Compilation for Aligned Agent Execution
浏览论文内容
中文总结 AI 辅助
研究针对大语言模型代理执行工作流指令时的未对齐问题，提出COVENANT架构，将指令转为工作流抽象语法树和控制流图，运行时控制器按要求检查提议并反馈。实验表明该方法大幅提升成功率、降低未对齐失败率，推动工作流可靠执行。
中文摘要 AI 辅助
大语言模型（LLM）代理越来越多地被赋予自然语言工作流指令（如零售支付政策），这些指令不仅指定要实现的结果，还规定了允许的步骤、分支和工具交互。然而，当这些指令作为提示上下文提供时，模型对过程选择和步骤执行仍有控制权。随着交互积累，代理可能会跳过必要步骤、采用无支持的分支，或使用无支持的参数或效果执行有效步骤，即工作流未对齐失败模式。在这项工作中，我们提出了COVENANT，一种用于工作流对齐代理执行的编译器和解释器架构。我们的关键见解是将工作流指令视为源程序而非提示。COVENANT将指令转换为工作流抽象语法树（WAST）并将其降低为工作流控制流图（WCFG）。运行时，控制器一次解释一个WCFG节点，在提交控制器状态或推进图之前，根据从指令中提取的要求检查每个提议，并返回诊断反馈以进行修复。为了评估COVENANT，我们使用了来自三个现有基准的120个案例，涵盖七个工作流场景。与最先进的LLM代理相比，COVENANT将基准成功率从50.00%提高到83.33%，并将工作流未对齐失败率从42.50%降低到15.83%（相对降低62.75%）。这些结果表明，COVENANT大大减轻了工作流未对齐问题，使LLM代理对齐从孤立的提示遵循转向复杂多步骤工作流的可靠执行。
英文摘要
Large language model (LLM) agents are increasingly entrusted with natural-language workflow instructions (e.g., retail-payment policies) that specify not only what outcome to achieve, but also which steps, branches, and tool interactions are permitted. When these instructions are supplied as prompt context, however, the model retains control over both procedure selection and step execution. As interactions accumulate, an agent can skip required steps, take unsupported branches, or execute a valid step with unsupported arguments or effects--a failure mode we call workflow misalignment. In this work, we propose COVENANT, a compiler-and-interpreter architecture for workflow-aligned agent execution. Our key insight is to treat workflow instructions as source programs rather than prompts. COVENANT converts the instructions into a workflow abstract syntax tree (WAST) and lowers it to a workflow control-flow graph (WCFG). At runtime, a controller interprets the WCFG one node at a time, checks each proposal against requirements extracted from the instructions before committing controller state or advancing the graph, and returns diagnostic feedback for repair. To evaluate COVENANT, we use 120 cases from three existing benchmarks, spanning seven workflow scenarios. Compared with state-of-the-art LLM agents, COVENANT improves benchmark success from 50.00% to 83.33% and reduces the workflow-misalignment failure rate from 42.50% to 15.83% (62.75% relative). These results show that COVENANT substantially mitigates workflow misalignment, moving LLM-agent alignment beyond isolated prompt following toward reliable execution of complex and multi-step workflows.
