让小模型在运行时自修工具流程，提升执行成功率。
Evoflux: Inference-Time Evolution of Executable Tool Workflows for Compact Agents
- 运行时通过演化搜索修复工具流程，动态调整结构与依赖。
- 在250个工具的测试中，执行成功率从3%提升至17%-24%。
- 适合资源受限场景下的智能代理，尤其需可靠工具调用的系统。
紧凑型语言模型可降低工具代理的成本、延迟和部署风险。然而，MCP风格的工具使用不仅需要孤立函数调用，还要求代理从实时目录中发现工具、满足模式约束、保持中间输出的依赖关系，并将最终响应基于已执行证据。小型规划器常生成看似合理但无法通过工具解析、参数验证、依赖追踪或执行的流程图。我们认为，这种失败模式难以通过小语料蒸馏解决：数百条教师轨迹可教会流程格式，却很少覆盖修复失败计划所需的恢复行为。为此提出Evoflux，一种推理时的进化搜索方法，将紧凑工具使用视为可执行流程的修复过程。它通过结构化编辑、执行反馈、自适应强度、元引导重设计和多样性剪枝，演化带类型的工作流图。在涵盖真实MCP服务器和250个工具的独立测试任务上，Evoflux将执行可行性从约3%提升至17%-24%。相比之下，相同数据上的SFT和SFT+DPO表现持平、欠优或低于零样本性能；ReAct虽达更高峰值，但方差大且消耗更多令牌。结果表明，在教师轨迹稀缺的情况下，基于执行的搜索更具可靠性。
原文摘要 · Abstract (English)
Compact language models (LMs) reduce cost, latency, and deployment risk for tool agents. Yet MCP-style tool use requires more than isolated function calling: an agent must discover tools from live catalogs, satisfy schemas, preserve dependencies across intermediate outputs, and ground final responses in executed evidence. Small planners often generate plausible workflow graphs that fail under tool resolution, parameter validation, dependency tracking, or execution. We argue that this failure mode is poorly handled by small-corpus distillation. A few hundred teacher traces can teach workflow format, but rarely cover the recovery behavior needed to repair failed plans over changing tool catalogs. We introduce Evoflux, an inference-time evolutionary search method that treats compact tool use as the repair of executable tool workflows. It evolves typed workflow graphs through structured edits, execution feedback, adaptive intensity, meta-guided redesign, and diversity pruning. On held-out MCP-Bench tasks spanning live MCP servers and 250 tools, Evoflux raises execution feasibility from roughly 3% to 17-24% across small planners. In contrast, SFT and SFT+DPO on the same search-mined data match, underperform, or collapse below zero-shot performance; ReAct reaches higher peaks, but with higher variance and token cost. These results show that execution-grounded search is more reliable under scarce teacher-trace budgets.
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。