1
00:00:00,320 --> 00:00:06,040
Imagine a profile form with a text field, a Save button, and a
working stored-value update.
设想一个资料表单，有文本框、保存按钮和正常的存储更新。

2
00:00:06,780 --> 00:00:12,060
The confirmation message never appears, so the user cannot tell
whether the requested action finished.
但确认消息始终不出现，用户无法判断操作是否完成。

3
00:00:12,700 --> 00:00:17,880
A repair could add that message, yet accidentally remove the
already working Save behavior.
修复可能补上消息，却意外破坏原本正常的保存行为。

4
00:00:18,560 --> 00:00:22,780
A prettier screenshot would not reveal that regression without
trying the control again.
如果不再次操作控件，更好看的截图无法揭示这种退化。

5
00:00:23,640 --> 00:00:28,600
This is an original schematic example, with no real user data or
measured repair outcome.
这是原创示意，不含真实用户数据或测量的修复结果。

6
00:00:29,240 --> 00:00:35,940
It introduces WebWorld's central question: what evidence should
allow a proposed web-code repair to survive?
它引出核心问题：什么证据能让网页代码修复被保留下来？

7
00:00:36,560 --> 00:00:43,620
WebWorld treats the browser as an executable simulator of how an
HTML artifact behaves under user actions.
WebWorld 把浏览器视为用户操作下执行 HTML 行为的模拟器。

8
00:00:44,500 --> 00:00:50,180
The runtime renders the current page, applies interaction
probes, and records evidence of what happened.
运行时渲染页面、执行交互探测，并记录发生的情况。

9
00:00:51,000 --> 00:00:58,000
That evidence includes screenshots, document state, console
information, probe outcomes, and an action trace.
证据包括截图、文档状态、控制台信息、探测结果和操作轨迹。

10
00:00:59,020 --> 00:01:05,640
A vision-language model reads those observations and proposes a
critique, rather than certifying its own repair.
视觉语言模型读取观察结果并提出批评，不自行认证修复。

11
00:01:06,540 --> 00:01:11,340
A planner converts the critique into something the browser can
check after the candidate edit.
规划器把批评转成候选编辑之后浏览器可检查的条件。

12
00:01:12,220 --> 00:01:18,800
Here, world model means the web execution platform; it does not
mean a physical-world prediction model.
这里的世界模型指网页执行平台，不是物理世界预测模型。

13
00:01:19,380 --> 00:01:25,780
An interaction contract states what must improve, how to test
it, and what existing behavior must survive.
交互契约规定改进目标、检查方式，以及必须保留的行为。

14
00:01:26,580 --> 00:01:31,860
In our form example, the target is a visible confirmation after
pressing the Save button.
在表单示例中，目标是点击保存之后出现可见确认。

15
00:01:32,520 --> 00:01:37,920
The replay specifies a stable control, an input action, and the
state to inspect afterward.
回放规定稳定控件、输入操作，以及之后检查的状态。

16
00:01:38,740 --> 00:01:45,680
The preserve set carries previously verified predicates, so
fixing one target cannot silently erase prior progress.
保留集合携带已验证条件，避免修复目标时抹去之前进展。

17
00:01:46,560 --> 00:01:51,180
For a local issue, the contract also limits the edit's impact to
the relevant region.
对于局部问题，契约也限制编辑影响的区域。

18
00:01:51,800 --> 00:01:58,000
These are bounded claims about specified interactions, rather
than a promise that every possible action will work.
这些是针对指定交互的有限断言，不保证所有操作都正常。

19
00:01:58,827 --> 00:02:03,707
A typed repair skill produces a constrained patch inside the
contract's impact scope.
类型化修复技能在契约影响范围内生成受约束的补丁。

20
00:02:04,627 --> 00:02:10,707
Candidate filters can discard unrelated, duplicate, or
preservation-risk edits before they reach the browser.
候选过滤器可提前丢弃无关、重复或有保留风险的编辑。

21
00:02:11,687 --> 00:02:15,227
That narrows the search, but it does not establish that the page
now works.
这缩小了搜索范围，但不能证明页面已经正常工作。

22
00:02:16,087 --> 00:02:21,247
The candidate must be executed again, with evidence tied to its
contract and recorded proof level.
候选需再次执行，证据绑定到契约和记录的证明级别。

23
00:02:22,147 --> 00:02:28,347
For the interactive form, replay checks the new confirmation and
rechecks the previously working Save behavior.
交互表单回放检查新确认，也重新检查原有保存行为。

24
00:02:29,107 --> 00:02:34,867
If either fails, our illustrative repair should remain a
rejected candidate rather than become the next baseline.
任一检查失败，示例修复就应保持拒绝状态，而非新基线。

25
00:02:35,560 --> 00:02:43,060
WebWorld distinguishes safe runtime steps from
certificate-backed quality steps that can be exported for
training.
WebWorld 区分安全运行步骤和可导出训练的认证质量步骤。

26
00:02:44,040 --> 00:02:49,360
A probe run, partial fix, or rejected edit can still help decide
what to try next.
探测、局部修复或被拒编辑仍可帮助决定下一次尝试。

27
00:02:50,080 --> 00:02:54,640
But those steps do not enter the supervised training pool simply
because the agent was active.
但智能体进行过操作，并不意味着这些步骤能进入监督训练池。

28
00:02:55,620 --> 00:03:02,480
Only admitted transitions carry the artifact, the failure it
addressed, and the evidence that justified acceptance.
被接纳的转移携带产物、解决的失败，以及支持接受的证据。

29
00:03:03,700 --> 00:03:10,080
Previously verified behavior becomes part of later preservation
checks, creating the proposed quality ratchet.
已验证行为进入后续保留检查，形成所提出的质量棘轮。

30
00:03:11,020 --> 00:03:15,140
The learning example therefore describes verified progress
between states,
因此学习样本描述状态之间的验证进展，

31
00:03:15,620 --> 00:03:18,200
rather than only a final preferred screenshot.
而不只是偏好的最终截图。

32
00:03:18,840 --> 00:03:22,140
The experiments compare raw HTML supervision
实验比较原始 HTML 监督，

33
00:03:22,200 --> 00:03:27,020
with supervision selected from thirty-two thousand eight hundred
accepted transitions.
与三万二千八百条接纳转移中选择的监督。

34
00:03:27,920 --> 00:03:34,600
Within each model size, the backbone family, prompt template,
optimizer, and evaluation runner remain fixed.
在每种模型规模内，骨干系列、提示、优化器和评估运行器固定。

35
00:03:35,440 --> 00:03:39,900
This makes the source of supervision the central difference in
the matched training comparison.
因此，监督来源是匹配训练比较中的主要差别。

36
00:03:41,220 --> 00:03:47,560
HTMLBench four hundred is the primary evaluation, with
MiniAppBench validation used to study transfer.
HTMLBench-400 是主要评估，MiniAppBench 验证集用于研究迁移。

37
00:03:48,640 --> 00:03:53,760
The construction pipeline reports separation from benchmark
prompts and their deterministic test cases.
构建流程报告与基准提示和确定性测试用例分离。

38
00:03:54,740 --> 00:04:01,240
The following chart compares the two
twenty-seven-billion-parameter rows, not unrelated
frontier-model configurations.
下一图比较两个二百七十亿参数行，不比较无关前沿模型配置。

39
00:04:01,780 --> 00:04:09,040
The raw-supervision twenty-seven-billion-parameter row reports
an HTMLBench composite score of forty-seven point four.
原始监督的二百七十亿参数行报告 HTMLBench 综合分四十七点四。

40
00:04:09,680 --> 00:04:14,740
The WebWorld row at the same model size reports fifty-two point
seven under that benchmark.
相同规模的 WebWorld 行在该基准上报告五十二点七。

41
00:04:15,340 --> 00:04:19,520
That is a gain of five point three score points in this matched
training setup.
在这一匹配训练设置中，综合分提高五点三分。

42
00:04:20,280 --> 00:04:24,880
The composite score is distinct from the separately reported
test-case pass percentage.
综合分与单独报告的测试用例通过百分比不同。

43
00:04:25,520 --> 00:04:31,280
The same model-size pair improves the MiniAppBench validation
average by fourteen point nine points.
同规模模型对的 MiniAppBench 验证平均分提高十四点九分。

44
00:04:32,120 --> 00:04:35,680
These results concern the reported interactive HTML benchmarks,
这些结果对应报告的交互 HTML 基准，

45
00:04:36,060 --> 00:04:38,660
rather than a guarantee of correctness for every website.
不保证每个网站都正确。

46
00:04:39,240 --> 00:04:45,800
A separate nine-billion-parameter ablation keeps critique and
repair routing while loosening the certificate gate.
独立九十亿参数消融保留批评与修复路由，放宽认证门控。

47
00:04:46,660 --> 00:04:52,400
Without the certificate, its HTMLBench gain over Raw is only
zero point four score points.
没有证书时，HTMLBench 相对 Raw 仅提高零点四分。

48
00:04:53,420 --> 00:04:58,680
The full WebWorld admission rule gives four point eight points
over Raw at that model size.
完整接纳规则在同规模模型上相对 Raw 提高四点八分。

49
00:04:59,720 --> 00:05:04,480
The experiment therefore supports more than the value of asking
a model to critique its own page.
因此实验支持的不只是让模型批评自身页面的价值。

50
00:05:05,100 --> 00:05:10,800
It tests whether the loop's selected training transitions are
backed by the specified verification rule.
它检验循环选择的训练转移是否由指定验证规则支持。

51
00:05:11,760 --> 00:05:17,440
The effect size belongs to this matched ablation; it should not
be generalized to every browser-checking system.
效应大小属于这一匹配消融，不应推广到所有浏览器检查系统。

52
00:05:18,087 --> 00:05:24,847
Return to our original form: a confirmation matters, but so does
the Save action that worked before.
回到原创表单：确认消息重要，原本正常的保存操作也重要。

53
00:05:25,527 --> 00:05:31,407
A candidate change should be described as a concrete target,
then tested under the relevant interactions.
候选变化应描述为具体目标，再在相关交互下测试。

54
00:05:32,107 --> 00:05:38,907
The working parts become explicit preservation obligations,
rather than assumptions hidden behind a polished screenshot.
正常部分成为明确保留要求，不再是漂亮截图背后的隐含假设。

55
00:05:39,867 --> 00:05:45,147
WebWorld makes that distinction part of both the repair loop and
the training-data admission rule.
WebWorld 把这种区分加入修复循环和训练数据接纳规则。

56
00:05:45,867 --> 00:05:52,327
A proposed improvement is a hypothesis; the recorded contract
evidence determines whether the system should keep it.
提出的改进是假设，记录的契约证据决定是否保留它。

57
00:05:53,407 --> 00:06:00,147
That is the idea to carry forward: self-improvement needs a
definition of progress that survives execution.
值得记住的是：自我改进需要一种经得起执行的进展定义。