WEBVTT

1
00:00:00.340 --> 00:00:08.280
Imagine four analysis stages: import records, apply a required
transformation, build a chart, and export.
设想四阶段分析请求：导入记录、必要转换、制图和导出。

2
00:00:09.400 --> 00:00:14.760
The chart may look plausible even when the agent skipped the
transformation specified in the request.
即使跳过指定转换，图表仍可能看起来合理。

3
00:00:15.460 --> 00:00:20.700
An export can therefore succeed as a software action while
delivering the wrong underlying result.
导出操作因此可能成功，但交付底层结果仍然错误。

4
00:00:21.900 --> 00:00:26.660
The stages are connected by their input state, not simply by the
order of visible windows.
阶段由输入状态相连，不只是屏幕窗口的先后顺序。

5
00:00:27.320 --> 00:00:31.680
This is a fictional short workflow, not one of the benchmark's
measured software tasks.
这是虚构短流程，不是基准中测量的软件任务。

6
00:00:32.259 --> 00:00:38.220
It introduces the long-horizon continuity that Workflow-GYM asks
computer-use agents to maintain.
它引出 Workflow-GYM 要求电脑操作智能体保持的长程连续性。

7
00:00:38.840 --> 00:00:47.200
Workflow-GYM evaluates complete computer-use workflows that
require professional software and domain-specific knowledge.
Workflow-GYM 评估需要专业软件与领域知识的完整电脑流程。

8
00:00:48.020 --> 00:00:55.020
Its pinned source contains three hundred thirty-eight tasks
across six domains and twenty-three finer subdomains.
固定版本含三百三十八个任务、六领域和二十三个子领域。

9
00:00:55.860 --> 00:01:02.620
The tasks require thirty to one hundred ten annotated
operations, rather than a handful of isolated clicks.
任务需三十到一百一十个标注操作，不是少量孤立点击。

10
00:01:03.460 --> 00:01:10.659
This creates dependencies between configuration, intermediate
state, and the final file or interface outcome.
这让配置、中间状态与最终文件或界面结果形成依赖。

11
00:01:11.580 --> 00:01:16.620
A locally reasonable action can still be wrong when the earlier
workflow state is incorrect.
当之前流程状态有误，局部合理操作仍可能错误。

12
00:01:17.240 --> 00:01:23.680
The benchmark therefore makes sustained execution part of what
it measures, alongside individual interface actions.
因此基准同时测量持续执行能力与单独界面操作。

13
00:01:24.300 --> 00:01:32.300
The environment design separates reusable software-level virtual
machines from lightweight task-specific initialization.
环境设计将可复用软件级虚拟机与轻量任务初始化分开。

14
00:01:33.540 --> 00:01:39.860
Fifty-six base virtual machines provide installed professional
software and its controlled version configuration.
五十六个基础虚拟机提供已安装专业软件和受控版本配置。

15
00:01:40.760 --> 00:01:47.540
For a task, input files, templates, or settings are injected
through defined initialization procedures.
每个任务按定义的初始化过程注入输入文件、模板或设置。

16
00:01:48.460 --> 00:01:54.520
The software starts ready for interaction, but the requested
task operations have not already been performed.
软件已准备交互，但请求的任务操作尚未执行。

17
00:01:55.540 --> 00:02:02.140
Irrelevant login prerequisites are removed, allowing the
evaluation to focus on the specified professional workflow.
无关登录前置条件被移除，使评估聚焦指定专业流程。

18
00:02:03.300 --> 00:02:09.380
In our diagram, these are generic workspaces, not credentials or
environments belonging to real individuals.
图中是通用工作区，不含真实个人凭据或环境。

19
00:02:10.020 --> 00:02:17.040
Each task begins with a self-contained instruction describing
its objective, required software, and final outcome.
每任务以自包含指令说明目标、所需软件与最终结果。

20
00:02:17.640 --> 00:02:22.540
The outcome may be a produced artifact or a specified final
graphical-interface state.
结果可以是生成的产物，也可以是指定最终图形界面状态。

21
00:02:23.740 --> 00:02:30.280
Experts provide atomic procedures and execute them during
validation to check that the task can be completed.
专家提供原子步骤并在验证中执行，以检查任务可完成。

22
00:02:31.040 --> 00:02:37.540
They revise problems in the environment, instructions, or
evaluation criteria when those prevent a valid solution.
若环境、指令或评价条件阻止有效解答，就进行修订。

23
00:02:38.440 --> 00:02:44.500
The main agent evaluation instead gives one initial instruction,
without step-by-step intermediate guidance.
主要智能体评估仅给一次初始指令，不提供逐步中间指导。

24
00:02:45.180 --> 00:02:50.220
The expert procedure defines the task; it is not the guidance
supplied during main evaluation runs.
专家步骤确定任务规范，不应误当每次运行中提供的提示。

25
00:02:50.860 --> 00:02:57.440
Difficulty bands follow the length of the annotated procedure:
thirty to forty-four operations form the easy group.
难度按标注步骤长度划分：三十至四十四操作属于简单组。

26
00:02:58.220 --> 00:03:04.520
Forty-five to sixty operations form the medium group, and
sixty-one to one hundred ten form the hard group.
四十五至六十为中等组，六十一至一百一十为困难组。

27
00:03:05.400 --> 00:03:10.920
These counts characterize task complexity, rather than
specifying the number of model interactions actually used.
这些计数描述任务复杂度，不规定模型实际使用的交互次数。

28
00:03:11.960 --> 00:03:18.200
Source analysis describes stage omission, error propagation, and
objective drift across these extended workflows.
源分析描述长流程中的阶段遗漏、错误传播和目标漂移。

29
00:03:19.240 --> 00:03:24.220
Skipping a required stage can leave later actions operating on a
state that was never properly constructed.
遗漏必要阶段会让后续操作基于从未正确建立的状态。

30
00:03:25.160 --> 00:03:30.760
Our original transformation example shows that problem without
borrowing any real software or benchmark trajectory.
原创转换示例展示这个问题，不借用真实软件或基准轨迹。

31
00:03:31.387 --> 00:03:37.547
The paper distinguishes workflow incompletion from reaching the
final stage with an incorrect result.
论文区分流程未完成与达到最终阶段但结果错误。

32
00:03:38.467 --> 00:03:43.827
An unfinished attempt may stop before export or leave the
software in an intermediate state.
未完成尝试可能在导出前停止，或停在软件中间状态。

33
00:03:44.767 --> 00:03:51.427
Another attempt can produce the expected file yet violate a
required configuration or omit a component.
另一尝试可生成所需文件，却违反配置要求或遗漏组件。

34
00:03:52.887 --> 00:03:59.307
Artifact tasks check whether the produced files satisfy the
specified requirements rather than merely existing.
产物任务检查文件是否满足要求，而不仅是存在。

35
00:04:00.267 --> 00:04:07.387
Non-artifact tasks use final-interface evidence and predefined
judging criteria to assess the requested state.
非产物任务使用最终界面证据与预定规则评价请求状态。

36
00:04:08.507 --> 00:04:14.227
The evaluation target remains the end-to-end outcome, not just
the plausibility of individual clicks.
评估目标仍是端到端结果，不只是每次点击看起来合理。

37
00:04:15.007 --> 00:04:20.467
Each model receives a single starting instruction and can use up
to four hundred interaction rounds.
每模型接收一次初始指令，最多四百个交互回合。

38
00:04:21.727 --> 00:04:27.807
The source uses model-specific agent frameworks, rather than one
identical harness for every model.
源实验使用模型特定框架，并非所有模型都用相同框架。

39
00:04:29.227 --> 00:04:36.147
Every task is attempted three times independently, producing one
thousand fourteen trials across the full task set.
每任务独立尝试三次，全任务集共一千零十四次试验。

40
00:04:37.587 --> 00:04:44.487
Average Pass pools success across those trials, while Pass at
three reports the separate three-attempt measure.
平均通过率汇总这些试验；Pass@3 报告另一个三次尝试指标。

41
00:04:46.047 --> 00:04:51.927
A greater chance of passing within repeated attempts is not the
same as reliable performance on each attempt.
多次尝试更可能成功，并不等于每次尝试都可靠。

42
00:04:53.167 --> 00:04:58.247
The following comparison keeps the same model row while
distinguishing those two reported metrics.
下一比较保留同一模型行，同时区分两个报告指标。

43
00:04:58.913 --> 00:05:05.573
For Gemini three point one Pro, the table reports Average Pass
of thirty point six seven percent.
Gemini-3.1-Pro 的平均通过率为百分之三十点六七。

44
00:05:06.213 --> 00:05:10.653
The same row reports Pass at three of forty-one point one two
percent.
同一行的 Pass@3 为百分之四十一点一二。

45
00:05:11.473 --> 00:05:17.753
They summarize the same evaluation protocol through different
success measures; neither value should replace the other.
它们用不同成功指标总结同一协议，不应相互替换。

46
00:05:18.813 --> 00:05:25.833
Kimi K two point six has a different Pass at three value, so
pairing it with Gemini's average would mix models.
Kimi-k2.6 的 Pass@3 不同，与 Gemini 平均值组合会混用模型。

47
00:05:26.933 --> 00:05:33.153
These figures belong to the three hundred thirty-eight curated
workflows and their model-specific execution settings.
这些数值属于三百三十八个整理流程及其模型特定执行设置。

48
00:05:33.693 --> 00:05:39.413
They should not be interpreted as a general success guarantee
for arbitrary professional software tasks.
它们不应被理解为任意专业软件任务的通用成功保证。

49
00:05:40.040 --> 00:05:46.740
Return to the original pipeline: the report depends on the
chart, and the chart depends on the transformation.
回到原创流程：报告依赖图表，图表依赖数据转换。

50
00:05:47.500 --> 00:05:54.000
A skipped stage can invalidate that chain while later software
actions still appear locally reasonable.
跳过一个阶段可破坏依赖链，后续操作却仍显得局部合理。

51
00:05:55.040 --> 00:06:01.760
Long-horizon computer use therefore requires maintaining the
intended state, not merely continuing to issue actions.
长程电脑操作因此需保持预期状态，不只是继续发送操作。

52
00:06:02.620 --> 00:06:08.520
Workflow-GYM tests continuity in controlled professional
software with defined end-to-end outcomes.
Workflow-GYM 在受控专业软件环境与明确端到端结果下揭示这种要求。

53
00:06:09.460 --> 00:06:15.440
The useful distinction is between reaching a final window and
producing the final state the user requested.
有用的区别是：到达最后窗口，还是生成用户请求的最终状态。

54
00:06:16.100 --> 00:06:22.000
A workflow succeeds when the important dependencies remain
correct all the way through the delivered outcome.
重要依赖一直保持正确直到交付结果，流程才成功。