> For AI agents: the complete documentation index is available at /zh/llms.txt, the full documentation bundle is available at /zh/llms-full.txt.
# Midscene MobileWorld Benchmark 测试报告
本文是 Midscene 针对 MobileWorld benchmark 的测试报告。本次测试仅覆盖 117 个 `GUI-ONLY` 任务，Midscene 取得了 **Pass\@1 78.63%（92/117）** 的结果。
:::info 关于 MobileWorld
[MobileWorld](https://github.com/Tongyi-MAI/MobileWorld) 是面向真实移动场景的 Android Agent benchmark，覆盖跨应用、长链路、用户交互和工具增强等任务，并通过可复现的 Android 环境验证 Agent 的执行结果。
:::
## 运行配置
| 字段 | 值 |
| --------------------------------- | -------------------------------------------------------------------------------------------- |
| 测试日期 | 2026-07-28 |
| Model Name | `Gemini-3.6-Flash` |
| Midscene version | `1.10.3` |
| Device | `DockerEmulator` |
| 任务范围 | 仅 `GUI-ONLY` 任务 |
| 任务数量 | 117 |
| `MIDSCENE_REPLANNING_CYCLE_LIMIT` | 50 |
| MobileWorld 工程 | MobileWorld 使用 Midscene benchmark 适配层，通过 Midscene RPC 执行 Agent，同时保留 MobileWorld 的任务初始化和验证流程。 |
| 验收规则 | 一个 MobileWorld validator 按任务意图做了校准，涉及的 case 见下方列表。 |
## 验证条件调整
本次 benchmark 对以下 MobileWorld 验证条件做了调整：
| 修改内容 | 涉及的 case |
| ------------------------------------------------------------------------------------------------------------------------------------------------ | ----------------------- |
| 旧 validator 要求 `created_at` 到 `expires_at` 必须间隔完整的 5 天。例如初始日期是 16 日，“五天后”是 21 日；但日期转换后的时分秒可能让实际间隔不足 5 × 24 小时，导致旧规则判定失败。现在改为接受超过 4 天且不超过 5 天的间隔。 | `MastodonNewFilterTask` |
## 报告文件
以下为本轮全部 117 个 `GUI-ONLY` 任务的详细报告。报告均为压缩后的自包含 HTML，点击“报告”会在新页面打开对应执行记录。
Round 1（117 份报告 · 92 PASS · 25 FAIL）
| # | 任务 | 状态 | 报告 |
| --- | ---------------------------------------- | -------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| 1 | AcceptMeetingTask | PASS | 报告 |
| 2 | AdjustBrightnessMaximumTask | PASS | 报告 |
| 3 | AdjustBrightnessMinimumTask | PASS | 报告 |
| 4 | AdjustFontIconMaximumTask | PASS | 报告 |
| 5 | AdjustFontIconMinimumTask | PASS | 报告 |
| 6 | BidFileRenameTask | PASS | 报告 |
| 7 | CVEmailTask | PASS | 报告 |
| 8 | CancelMeetingTask | PASS | 报告 |
| 9 | CartInfoNotificationTask | PASS | 报告 |
| 10 | CartManagementTask | FAIL | 报告 |
| 11 | ChangeWallpaperTask | PASS | 报告 |
| 12 | CheckCartPriceTask | PASS | 报告 |
| 13 | CheckConferenceAndSendSmsTask1 | PASS | 报告 |
| 14 | CheckConferenceAndSendSmsTask2 | FAIL | 报告 |
| 15 | CheckConferenceDurationTask | PASS | 报告 |
| 16 | CheckConferenceLocationTask | FAIL | 报告 |
| 17 | CheckDeduplicatedEventsTask | PASS | 报告 |
| 18 | CheckDepartTimeTask | PASS | 报告 |
| 19 | CheckEventTimeTask | PASS | 报告 |
| 20 | CheckGithubInfoTask | PASS | 报告 |
| 21 | CheckInterviewTimesTask | FAIL | 报告 |
| 22 | CheckInvoiceTask1 | PASS | 报告 |
| 23 | CheckInvoiceTask2 | PASS | 报告 |
| 24 | CheckInvoiceTask3 | PASS | 报告 |
| 25 | CheckPuchasedItem | PASS | 报告 |
| 26 | CheckRegistrationTask | PASS | 报告 |
| 27 | CheckSetMeetTimeTask | PASS | 报告 |
| 28 | ChromeSearchBeijingWeatherTask | PASS | 报告 |
| 29 | CloseFlightModeTask | PASS | 报告 |
| 30 | CountFileLinesTask | PASS | 报告 |
| 31 | DownloadSendReceiptTask | PASS | 报告 |
| 32 | GoogleMapsAlibabaPhoneContactTask | PASS | 报告 |
| 33 | GoogleMapsAlibabaSouthNeighborTask | PASS | 报告 |
| 34 | GraduationMassEmailTask | PASS | 报告 |
| 35 | InvoiceReceiptCopyAskUserTask | FAIL | 报告 |
| 36 | InvoiceReceiptCopyTask | PASS | 报告 |
| 37 | ItemCheckoutTask | PASS | 报告 |
| 38 | LocalFileManagementTask | PASS | 报告 |
| 39 | LocalFileManagementTask2 | PASS | 报告 |
| 40 | MastodonAddBookmarkTask | PASS | 报告 |
| 41 | MastodonAddFeaturedHashtagsTask | FAIL | 报告 |
| 42 | MastodonAdjustTootsTask | PASS | 报告 |
| 43 | MastodonCalendarMultiMemosTask | FAIL | 报告 |
| 44 | MastodonChangeHeaderTask | PASS | 报告 |
| 45 | MastodonChangeLanguageTask | FAIL | 报告 |
| 46 | MastodonConditionalFavoTask | PASS | 报告 |
| 47 | MastodonCreateListTask | PASS | 报告 |
| 48 | MastodonCreateMemoTask | PASS | 报告 |
| 49 | MastodonExportFollowsTask | PASS | 报告 |
| 50 | MastodonFavoriteTootsTask | PASS | 报告 |
| 51 | MastodonFilterLanguageTask | PASS | 报告 |
| 52 | MastodonFollowTask | PASS | 报告 |
| 53 | MastodonGetServerInfoTask | FAIL | 报告 |
| 54 | MastodonImportMutedUsersTask | PASS | 报告 |
| 55 | MastodonInviteTask | PASS | 报告 |
| 56 | MastodonMallPurchaseCommodityTask | PASS | 报告 |
| 57 | MastodonMallShareOrderTask | PASS | 报告 |
| 58 | MastodonManageHashtagsTask | PASS | 报告 |
| 59 | MastodonManageMultiListTask | PASS | 报告 |
| 60 | MastodonMattermostPostNoticeTask | PASS | 报告 |
| 61 | MastodonMultiInviteTask | PASS | 报告 |
| 62 | MastodonNewFilterTask | FAIL | 报告 |
| 63 | MastodonNewPostTask | PASS | 报告 |
| 64 | MastodonOpenAutomatedDeletionTask | PASS | 报告 |
| 65 | MastodonPinTootsTask | PASS | 报告 |
| 66 | MastodonPostEditedPhotoTask | PASS | 报告 |
| 67 | MastodonPostPollTask | PASS | 报告 |
| 68 | MastodonRemoveBookmarkTask | PASS | 报告 |
| 69 | MastodonReplyTask | PASS | 报告 |
| 70 | MastodonReportTask | PASS | 报告 |
| 71 | MastodonRevisePhotoAltTask | PASS | 报告 |
| 72 | MastodonRevisePollTask | PASS | 报告 |
| 73 | MastodonSavePhotosTask | PASS | 报告 |
| 74 | MastodonServerInfoReportTask | FAIL | 报告 |
| 75 | MastodonShareLocationTask | FAIL | 报告 |
| 76 | MastodonUnfollowTask | PASS | 报告 |
| 77 | MastodonUpdateContactsTask | PASS | 报告 |
| 78 | MattermostBudgetApprovalPipelineTask | PASS | 报告 |
| 79 | MattermostCreateChannelTask | PASS | 报告 |
| 80 | MattermostCustomerFeedbackAnalysisTask | PASS | 报告 |
| 81 | MattermostDeadlineReconciliationTask | PASS | 报告 |
| 82 | MattermostEmailTask | PASS | 报告 |
| 83 | MattermostIncidentEscalationTask | PASS | 报告 |
| 84 | MattermostProjectHandoverTask | FAIL | 报告 |
| 85 | MattermostProjectStatusReportTask | FAIL | 报告 |
| 86 | MattermostReadingGroupTask | FAIL | 报告 |
| 87 | MattermostReplyToMessageTask | PASS | 报告 |
| 88 | MattermostResourceConflictResolutionTask | FAIL | 报告 |
| 89 | MattermostSendFileTask | FAIL | 报告 |
| 90 | MattermostShiftCoverageTask | PASS | 报告 |
| 91 | MattermostTechnicalDebtTriageTask | PASS | 报告 |
| 92 | MattermostVisualInstructionResponseTask | FAIL | 报告 |
| 93 | OpenFlightModeTask | PASS | 报告 |
| 94 | PhotoManagementTask | FAIL | 报告 |
| 95 | ReadQwen3PaperTask1 | PASS | 报告 |
| 96 | ReadQwen3PaperTask2 | FAIL | 报告 |
| 97 | ReadQwen3PaperTask3 | FAIL | 报告 |
| 98 | ReadQwen3PaperTask4 | PASS | 报告 |
| 99 | ReadQwen3PaperTask5 | FAIL | 报告 |
| 100 | RecentTotalExpenseTask | PASS | 报告 |
| 101 | RequestCarpoolingTask | PASS | 报告 |
| 102 | ReviewPaperEmailTask | FAIL | 报告 |
| 103 | SMSManagement | PASS | 报告 |
| 104 | ScheduleCoffeeTimeViaSmsTask | PASS | 报告 |
| 105 | ScheduleLunchViaSmsTask | PASS | 报告 |
| 106 | SearchItemAndCheckoutTask | PASS | 报告 |
| 107 | SendFormsTask | PASS | 报告 |
| 108 | SendInterviewEmailTask | PASS | 报告 |
| 109 | SendInterviewInvitationTask | FAIL | 报告 |
| 110 | SendWaiverTask | PASS | 报告 |
| 111 | SetAlarmTask | PASS | 报告 |
| 112 | SharePhotosTask | PASS | 报告 |
| 113 | SuggestPaperTask | PASS | 报告 |
| 114 | SumFileLinesTask | PASS | 报告 |
| 115 | TakeSelfieTask | PASS | 报告 |
| 116 | TextArrivalTimeTask | FAIL | 报告 |
| 117 | ThanksgivingPrepTask | PASS | 报告 |