WebWorld：将浏览器作为自改进网页代码的世界模型

中文摘要 AI 辅助

基于视觉语言模型（VLM）的网页代码自改进存在结构性缺陷：提出修复方案的模型同时也是评判该方案的模型，而在该评判标准下的视觉合理性，与网页实际是否可用的关联度极低。该循环缺少一个VLM无法欺骗的对象，而浏览器正是这样的对象：它是HTML制品在用户操作下行为的确定性、可执行模拟器，本质上就是网页代码的世界模型。我们提出WebWorld，该接口让预训练VLM能自主与这个作为世界模型的浏览器交互，并决定哪些交互可作为监督信号。每一轮中，VLM生成一份批评意见，规划器将其编译为带类型的交互契约；浏览器会重新执行候选代码，仅当满足目标进展且保留所有先前验证过的功能时，才会签发合格证书；经认证的转换会累积为质量棘轮，这也是SFT（监督微调）导出时唯一可见的内容。在匹配训练下，WebWorld-27B在HTMLBench-400上较Raw-27B提升5.3个百分点，在MiniAppBench-Val上提升14.9个百分点，在交互式HTML生成上达到Kimi-K2.6、GPT-5.4等强前沿系统的水平。同等规模的消融实验表明，浏览器背书的准入机制是性能提升的关键：若无合格证书，匹配训练的9B模型的提升几乎消失。

英文摘要

VLM-driven self-improvement of web code has a structural flaw: the model that proposes the repair is the model that judges it, and visual plausibility under that judge is a poor proxy for whether the page actually works. What the loop is missing is a counterparty the VLM cannot fool, and the browser already is that counterparty: a deterministic, executable simulator of how an HTML artifact behaves under user actions, and in everything but name a world model for web code. We present WebWorld, the interface that lets a VLM prior interact with this browser-as-world-model autonomously and decides which interactions become supervision. Each round, the VLM emits a critique that the planner compiles into a typed interaction contract; the browser re-executes the candidate and issues an acceptance certificate only when both target progress and preservation of every previously verified capability hold; certified transitions accumulate as a quality ratchet that is the only thing the SFT export ever sees. Under matched training, WebWorld-27B improves Raw-27B by 5.3 points on HTMLBench-400 and 14.9 points on MiniAppBench-Val, and reaches the level of strong frontier systems such as Kimi-K2.6 and GPT-5.4 on interactive HTML generation. Equal-size ablations show that browser-backed admission carries the gain: without the certificate, the matched 9B lift nearly disappears.

发表机构

• Beihang University（北京航空航天大学）
• Shanghai Jiao Tong University（上海交通大学）
• IQuest Research（IQuest研究院）
• Langboat

机构由 AI 辅助整理，请以论文原文为准。

补充信息

备注

EMNLP Main Conference