AI 练手的考场，终于从题库变成了真实工位
大模型当「智能体」干活，靠的是把一堆工具串起来。但之前训练它的考场是假的：先编个数据库和场景，再倒推出工具和任务，工具接口对不上真的，任务也几乎不跨环境。这篇把顺序反过来：从真实 MCP 服务的执行日志里，用确定性规则挖出工具之间的依赖关系，按依赖把工具聚成一个个「环境」，再在环境里做带状态回滚的随机游走，生成任务和校验。结果：7 个前沿模型在它生成的更难版本上，分数最多掉 38.2 分；用它训练的小模型，平均分从 23 涨到 41，追平了顶级闭源模型。它不是你明天能用上的东西，但它是「AI 到底会不会干活」这个问题的更诚实考卷。
📄 原文摘要(英文)
Large language models are deployed as long-horizon agents that compose diverse tools across MCP services into executable workflows. Existing MCP environment-scaling methods are data-centric: they first synthesize a database and scenario, then derive tools and tasks from them. As a result, tool interfaces may not match real ones, tool compositions are not grounded in execution, and tasks rarely span more than one environment. We propose ToolSpace, a tool-centric framework that reverses this order. ToolSpace mines data, state, and retry dependencies from adjacent calls in execution traces of real MCP services and open agent corpora, using deterministic rules rather than LLM judgment, and clusters the resulting weighted tool-composition graph into environments. Databases are synthesized only for state-mutating environments, while query and stateless tools replay recorded returns. ToolSpace then performs state-aware random walks with rollback and decompiles each chain into a natural-language task with assertion-based checks, keeping only samples that replay successfully from the initial state. Applied to AutomationBench and tau3-Bench Banking, the pipeline produces harder Pro versions on which 7 frontier models drop by up to 38.2 points. Across seven long-horizon tool-use benchmarks, ToolSpace raises Qwen3.6-35B-A3B from 23.0 to 41.0 average score (43.2 with RL) and GLM-5.2 from 44.6 to 56.6, on par with Claude Opus 5. Ablations show that execution-witnessed edges clearly outperform LLM-judged ones, and performance keeps improving with data and environment scale.