StateM：通过Harness扩展（Harness Scaling），在Terminal-Bench 2.1上达到95.3%原始准确率——即一次15美元的前沿级运行

摘要

长时程智能体即使其底层模型能够解决各个组成步骤，也可能失败。它们可能丢失可变状态的踪迹，无法重新启用先前执行中的经验教训，跳过已知流程，或过早停止。我们押注于扩展执行框架（harness scaling）来改进智能体周围的执行系统，而不改变其模型权重。我们引入了StateM，一种智能体原生的运行时，它围绕持久状态、阶段局部上下文、受检转换、可恢复运行手册以及智能体和用户可共同检查的版本化流程实践来组织执行。在Terminal-Bench 2.1上，StateM将GPT-5.5 xhigh提升至92.1%，而参考值为83.1%，GPT-5.6 Sol Ultra为91.9%。该运行手册无需修改即可迁移到GPT-5.6。使用GPT-5.6 Sol xhigh，StateM在445次试验中达到了95.3%的原始准确率，并在全部89个任务上至少成功一次。冻结的配置将GPT-5.6 Luna从76.7%提升至85.4%，超过了84.9%的Sol xhigh参考值。使用相同的运行时、运行手册结构和黄金规则，不到38美元的适配投入就能使DeepSeek-V4 Flash在标准超时下从82.7%提升至88.1%，并在88个任务的公共核心子集上达到89.1%。仅扩展剩余的延迟敏感任务即可匹配所报告的88.8% GPT-5.6 Sol最高结果。最终评分的API使用量约为15美元，而GPT参考为574.68美元；DeepSeek总支出为52.22美元。在BusinessBench上，基于开发集构建的特定任务族运行手册在留出数据上分别带来0.55个宏平均点和1.34个微平均点的提升；两个机制匹配的任务族提升了10.04个百分点。当任务共享执行结构时，具体规则能够泛化，而控制方法论则广泛适用。StateM将选定的事后分析发现转化为持久的、可执行的先决条件与实践，通过有状态控制使习得的控制策略变得明确且可强制执行。代码见github.com/henryqin1997/statem。

English

Long-horizon agents can fail even when their underlying models can solve the constituent steps. They may lose track of mutable state, fail to reactivate lessons from earlier executions, skip known procedures, or stop prematurely. We bet on harness scaling to improve the execution system around an agent without changing its model weights. We introduce StateM, an agent-native runtime that organizes execution around durable states, phase-local context, checked transitions, recoverable runbooks, and versioned procedural practices that agents and users can inspect together. On Terminal-Bench 2.1, StateM raises GPT-5.5 xhigh to 92.1\%, versus 83.1\% reference and GPT-5.6 Sol Ultra at 91.9\%. The runbook transfers unchanged to GPT-5.6. With GPT-5.6 Sol xhigh, StateM reaches 95.3\% raw accuracy across 445 trials and succeeds on all 89 tasks at least once. The frozen profile raises GPT-5.6 Luna from 76.7 to 85.4\%, above the 84.9\% Sol xhigh reference. Using the same runtime, runbook structure, and golden rules, less than \38 of adaptation raises DeepSeek-V4 Flash from 82.7 to 88.1% under standard timeouts and to 89.1% on an 88-task common core. Extending only the remaining latency-sensitive task matches the reported 88.8% GPT-5.6 Sol max result. Final-score API usage is about 15 versus \574.68 for the GPT reference; total DeepSeek expenditure is 52.22. On BusinessBench, family-specific runbooks built on development sets yield held-out gains of 0.55 macro and 1.34 micro points; two mechanism-matched families improve by 10.04 points. Concrete rules generalize when tasks share execution structure, while the control methodology applies broadly. StateM turns selected postmortem findings into persistent, executable preconditions and practices, making learned controls explicit and enforceable through stateful controls. Code at github.com/henryqin1997/statem.