1
00:00:00,260 --> 00:00:06,420
A deployed agent may receive a task instruction without a
ready-made skill library or reliable practice tests.
部署后的智能体可能只收到任务指令，却没有现成的技能库或可靠的练习测试。

2
00:00:07,220 --> 00:00:12,480
OpenSkill asks how that agent can build both a useful procedure
and a way to assess it.
OpenSkill 研究这种智能体如何同时构建有用的操作流程，以及评估该流程的方法。

3
00:00:13,140 --> 00:00:18,400
Its adaptation lives in external skill artifacts, leaving the
underlying model weights unchanged.
它的适应能力体现在外部技能产物中，而底层模型权重保持不变。

4
00:00:19,040 --> 00:00:24,440
The pipeline uses public documentation, repositories, and other
open-world resources as learning material.
这个流程把公开文档、代码仓库和其他开放世界资源作为学习材料。

5
00:00:25,120 --> 00:00:31,080
The central design is to retrieve knowledge for the skill and
verifiable anchors for its tests as separate products.
核心设计是分别获取用于技能的知识，以及用于测试的可验证依据。

6
00:00:31,780 --> 00:00:36,380
We will follow an original dataset-report example, then examine
what the measured outcomes support.
我们先跟着一个原创数据集报告示例，再查看实测结果能够支持什么结论。

7
00:00:37,087 --> 00:00:42,187
Imagine a task that asks for a report from a public dataset,
with a specified output format.
想象一项任务，要求根据公开数据集生成报告，并使用指定的输出格式。

8
00:00:42,827 --> 00:00:48,667
Knowledge retrieval can supply loading procedures, documented
library behavior, and useful domain conventions.
知识检索可以提供加载流程、文档中的库行为，以及有用的领域惯例。

9
00:00:49,347 --> 00:00:52,327
A separate retrieval product supplies verification anchors:
另一份检索产物提供验证依据：

10
00:00:52,907 --> 00:00:57,687
documented reference values, known dataset invariants, or
expected formats.
文档参考值、已知数据集不变量，或预期格式。

11
00:00:58,327 --> 00:01:04,647
In our illustration, a known row count and a documented table
schema give the checker concrete assertions to execute.
在我们的示例中，已知行数和文档中的表格结构，为检查器提供了可执行的具体断言。

12
00:01:05,367 --> 00:01:11,387
Those checks assess properties of an output, rather than simply
asking the model whether its own answer looks good.
这些检查评估输出的属性，而不是只问模型自己的回答看起来是否不错。

13
00:01:12,147 --> 00:01:18,267
The example contains no hidden benchmark answer. Its purpose is
to show how an external fact becomes a test.
这个示例没有包含隐藏基准答案；它用于展示外部事实如何转化为测试。

14
00:01:18,913 --> 00:01:24,393
The retrieved knowledge becomes a structured plan describing the
skill architecture, procedures, and domain rules.
检索到的知识会转化为结构化计划，描述技能架构、流程和领域规则。

15
00:01:24,933 --> 00:01:30,173
The base agent drafts one to four skills, as specified by that
plan, and refines them together in one session.
基础智能体按照计划起草一到四项技能，并在同一个会话中共同完善它们。

16
00:01:30,713 --> 00:01:35,753
Because the skill is an external artifact, its procedure remains
readable and can be passed to another agent.
因为技能是外部产物，其流程仍然可读，也可以传递给另一个智能体。

17
00:01:36,433 --> 00:01:41,093
In the report illustration, the skill would connect loading the
input, preserving relevant information,
报告示例中，技能连接输入加载与相关信息保留，

18
00:01:41,273 --> 00:01:42,653
and writing the required output.
以及写入所需输出。

19
00:01:43,553 --> 00:01:49,213
The verifier is an isolated language-model session that produces
deterministic executable tests from the anchors.
验证器是隔离的语言模型会话，依据验证材料生成确定性的可执行测试。

20
00:01:49,733 --> 00:01:52,173
A separate session isolates the creator's conversation;
独立会话隔离创建者的对话；

21
00:01:52,533 --> 00:01:55,413
it does not make proxy tests identical to the hidden target
tests.
但不会让代理测试与隐藏目标测试完全相同。

22
00:01:55,900 --> 00:01:59,680
The current skill is executed and evaluated against the
self-built virtual suite.
当前技能会被执行，并接受自建虚拟测试集的评估。

23
00:02:00,260 --> 00:02:05,320
A failing assertion generates a diagnostic: what failed, why it
failed, and what should be revised.
断言失败后会产生诊断信息：哪里失败、为何失败，以及应该修改什么。

24
00:02:06,240 --> 00:02:11,320
If the issue is a knowledge gap rather than an implementation
bug, the pipeline performs targeted retrieval.
如果问题是知识缺口，而不是实现错误，流程会进行有针对性的检索。

25
00:02:12,160 --> 00:02:16,060
The default loop stops when the virtual suite passes or after
three refinement rounds,
默认在虚拟测试通过后，或三轮改进后停止，

26
00:02:16,180 --> 00:02:17,820
with additional budget and stall controls.
另有预算与停滞控制。

27
00:02:18,640 --> 00:02:21,740
The final deployed skill is the last refined version at
termination,
最终部署技能是停止时最后修订的版本，

28
00:02:21,980 --> 00:02:24,500
rather than a best-of-many snapshot selected afterward.
不是之后从多个版本中挑选的最优快照。

29
00:02:25,300 --> 00:02:30,880
Virtual success is still a proxy. The later target evaluation
measures whether that practice produced useful behavior.
虚拟测试成功仍然只是代理指标；后续目标评估会测量这些练习是否产生了有用行为。

30
00:02:31,427 --> 00:02:36,147
The architecture hides ground-truth test definitions and
reference solutions during skill construction.
该架构在构建技能时隐藏真值测试定义和参考解答。

31
00:02:36,767 --> 00:02:41,327
Retrieval queries also filter benchmark names and identifiers
that could lead to hidden test material.
检索查询也会过滤基准名称，以及可能指向隐藏测试材料的标识符。

32
00:02:42,147 --> 00:02:47,507
The verifier receives the task environment and independent
anchors, with a separate session from the skill creator.
验证器接收任务环境和独立验证依据，并与技能创建者使用不同会话。

33
00:02:48,347 --> 00:02:52,527
Appendix F also describes a single ground-truth-derived
pass-or-fail bit
附录 F 也描述了一个来自真值的通过或失败比特，

34
00:02:52,567 --> 00:02:54,747
when the verifier is asked to write deeper tests.
用于再次要求验证器编写更深入测试时。

35
00:02:55,367 --> 00:03:00,747
It states that test contents, assertion details, and expected
values are not exposed in that invocation.
论文指出，这次调用不会暴露测试内容、断言细节或预期值。

36
00:03:01,467 --> 00:03:03,547
The boundary therefore concerns hidden contents;
因此边界主要针对隐藏内容；

37
00:03:04,107 --> 00:03:08,407
the description does not justify saying every possible
ground-truth-derived signal is absent.
该描述不足以说明所有来自真值的信号都不存在。

38
00:03:09,220 --> 00:03:13,840
The paper compares virtual-verifier decisions with ground-truth
outcomes on eighty-four tasks.
论文在八十四项任务上，对照虚拟验证器的决定与真值结果。

39
00:03:14,400 --> 00:03:20,220
Reported agreement is sixty point seven percent, with precision
fifty-six point nine and recall eighty point five.
报告的一致率为百分之六十点七，精确率为百分之五十六点九，召回率为百分之八十点五。

40
00:03:20,800 --> 00:03:24,800
Here, the comparison distinguishes whether a task receives
positive ground-truth reward,
此对比区分任务是否获得正的真值奖励，

41
00:03:25,180 --> 00:03:27,360
rather than certifying every hidden requirement.
不是认证每个隐藏要求都已满足。

42
00:03:28,000 --> 00:03:32,180
In a separate fifteen-task sample, a language-model judge
matches virtual tests
独立十五任务样本中，语言模型评判器匹配虚拟测试，

43
00:03:32,260 --> 00:03:34,420
to the intended purposes of hidden tests.
匹配对象是隐藏测试的预期目的。

44
00:03:34,960 --> 00:03:39,660
It finds one hundred twenty of one hundred thirty-five intents
covered, or eighty-eight point nine percent.
它发现一百三十五个测试目的中的一百二十个被覆盖，即百分之八十八点九。

45
00:03:40,320 --> 00:03:42,780
Intent coverage and outcome agreement are different
measurements.
目的覆盖率与结果一致率是不同测量。

46
00:03:43,220 --> 00:03:46,160
Neither means a proxy pass is always a target-task pass.
两者都不表示代理通过必然是目标任务通过。

47
00:03:46,820 --> 00:03:50,660
The main SkillsBench comparison covers eleven domains and two
target agents:
主要 SkillsBench 比较覆盖十一个领域与两个目标智能体：

48
00:03:51,200 --> 00:03:52,780
Opus four point six with Claude Code,
Claude Code 中的 Opus 4.6，

49
00:03:53,220 --> 00:03:55,020
and GPT-five point two with Codex.
以及 Codex 中的 GPT-5.2。

50
00:03:55,620 --> 00:04:00,480
For Opus, the reported overall pass rate rises from twenty-five
point five without skills
Opus 的总体通过率从无技能时百分之二十五点五，

51
00:04:00,860 --> 00:04:03,120
to forty-three point six with OpenSkill.
升至 OpenSkill 下的百分之四十三点六。

52
00:04:03,700 --> 00:04:08,000
For GPT-five point two, it rises from twenty-five to forty-two
point one percent.
对于 GPT-5.2，该数值从百分之二十五升至百分之四十二点一。

53
00:04:08,900 --> 00:04:14,980
These outcomes belong to the evaluated agent, framework, and
task setup, not a model-only causal comparison.
这些结果对应被评估的智能体、框架和任务配置，而不是只针对模型的因果对比。

54
00:04:15,700 --> 00:04:19,920
SocialMaze and ScienceWorld provide two additional task types in
the paper's evaluation.
SocialMaze 和 ScienceWorld 在论文评估中提供了另外两种任务类型。

55
00:04:20,780 --> 00:04:23,640
The external skill artifact is frozen before target deployment,
外部技能产物在目标部署之前被冻结，

56
00:04:24,080 --> 00:04:26,780
and target tests determine the resulting measured outcome.
目标测试决定最终测得结果。

57
00:04:27,267 --> 00:04:33,107
OpenSkill couples open-world learning with externally anchored
practice, while leaving model weights fixed.
OpenSkill 把开放世界学习与有外部依据的练习结合起来，同时保持模型权重不变。

58
00:04:33,667 --> 00:04:38,047
The report illustration shows the essential split: learn how to
produce the artifact,
报告示例展示关键分工：学习如何生成产物，

59
00:04:38,227 --> 00:04:40,807
and independently specify properties worth checking.
并独立规定值得检查的属性。

60
00:04:41,367 --> 00:04:47,127
Diagnostics revise the procedure, and frozen skills carry that
learned procedure into later target execution.
诊断信息修订流程，冻结后的技能把学到的流程带入后续目标执行。

61
00:04:47,787 --> 00:04:53,567
The paper also deploys Opus-generated skills to four other
models without retraining or regenerating those skills.
论文还把 Opus 生成的技能部署给其他四个模型，而不重新训练或重新生成这些技能。

62
00:04:54,347 --> 00:04:59,667
Useful adaptation therefore needs inspectable skills, grounded
tests, and measured target outcomes,
有用适应需要可检查技能、有依据测试与测得目标结果，

63
00:05:00,067 --> 00:05:02,067
with the information boundary stated precisely.
同时准确说明信息边界。

64
00:05:02,627 --> 00:05:06,547
The practical question is whether the skill learned from
available resources helps
实际问题是可用资源中学到的技能是否有帮助，

65
00:05:06,587 --> 00:05:08,867
when the real task finally has to be executed.
特别在真正需要执行任务时。