2026.09.28 · 论文解读 [2026-08-03] LongHorizon-Harness: Advancing Long-Horizon Agents for Real-World Tasks 拆解 LongHorizon-Harness 如何单独保存任务进度、每轮重新启动执行会话并独立核验结果；结合真实失败案例，说明评分、成本与实验限制。
