SPIN让工业级大模型规划更可靠，减少无效操作。
SPIN: Structural LLM Planning via Iterative Navigation for Industrial Tasks
- 用有向无环图约束规划过程，确保流程结构合法
- 执行前验证并修复计划，任务量减少41%，成功率提升至70.6%
- 适合需要稳定执行的工业自动化场景
工业级大模型代理系统常将规划与执行分离，但大模型规划器生成的流程常存在结构错误或冗余，导致失败和不必要的工具调用成本。我们提出 exttt{SPIN}，一种结合验证后有向无环图（DAG）规划与前缀式执行控制的规划封装器。 exttt{SPIN} 通过 exttt{\_validate\_plan\_text} 和修复提示严格遵守 DAG 合约，在下游执行前生成可执行计划，并增量评估 DAG 前缀，当当前前缀足以回答问题时立即停止。在 AssetOpsBench 上，261 个场景中， exttt{SPIN} 将执行任务数从 1061 减少至 623，完成率从 0.638 提升至 0.706，每轮工具调用从 11.81 次降至 6.82 次。在 MCP Bench 上，该封装器同样提升了 GPT OSS1 与 Llama 4 Maverick 的规划、对齐与依赖相关得分。
原文摘要 · Abstract (English)
Industrial LLM agent systems often separate planning from execution, yet LLM planners frequently produce structurally invalid or unnecessarily long workflows, leading to brittle failures and avoidable tool and API cost. We propose \texttt{SPIN}, a planning wrapper that combines validated Directed Acyclic Graph (DAG) planning with prefix based execution control. \texttt{SPIN} enforces a strict DAG contract through \texttt{\_validate\_plan\_text} and repair prompting, producing executable plans before downstream execution, and then evaluates DAG prefixes incrementally to stop when the current prefix is sufficient to answer the query. On AssetOpsBench, across 261 scenarios, \texttt{SPIN} reduces executed tasks from 1061 to 623 and improves \emph{Accomplished} from 0.638 to 0.706, while reducing tool calls from 11.81 to 6.82 per run. On MCP Bench, the same wrapper improves planning, grounding, and dependency related scores for both GPT OSS1 and Llama 4 Maverick.
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。