# Catalyzing Interactive HTML Generation Capabilities via High-Fidelity Reinforcement Learning

DevFeed: [Catalyzing Interactive HTML Generation Capabilities via High-Fidelity Reinforcement Learning](<https://devfeed.tech/articles/catalyzing-interactive-html-generation-capabilities-via-high-fidelity-reinforcement-learning-117569.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.05151>)

Author: Xujia Wang, Yixuan Liu, Shaohan Huang, Furu Wei, Lei Hou, Bin Xu, Juanzi Li

Published: 2026-11-04T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [Interactive AI Systems](<https://devfeed.tech/topics/interactive-ai-systems.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>)

Tags: [browser](<https://devfeed.tech/tags/browser.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [games](<https://devfeed.tech/tags/games.md>), [generation](<https://devfeed.tech/tags/generation.md>), [grounding](<https://devfeed.tech/tags/grounding.md>), [html](<https://devfeed.tech/tags/html.md>), [interactive](<https://devfeed.tech/tags/interactive.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [vlm](<https://devfeed.tech/tags/vlm.md>)

## AI overview

WebCatalyst uses browser-interaction reinforcement learning to improve the functionality and visual quality of generated HTML. Its three components are a game-level GUI assessor for sustained interaction evaluation, Visual Anchor to mitigate vision-language-model reward hacking, and Action-Trace Credit to ground token-level credit in runtime traces. Trained on fewer than 1,000 requests, mostly for games, it raises Qwen3.6-35B-A3B's Cookie-Bench Mini Games score from 68.8 to 75.3 and WebGen-Bench accuracy from 48.8 to 56.9; the authors also report a better average rank than Claude Opus 4.7 across evaluated metrics. Their analysis attributes reduced aesthetic degradation to Visual Anchor and faster functional improvement to runtime-grounded credit assignment.

## Source excerpt

arXiv:2611.05151v1 Announce Type: new Abstract: Generating HTML from natural-language specifications is an important application of code large language models (LLMs). Despite producing renderable pages, current models often fail to ensure functionality and visual quality during actual interaction. Existing methods typically target task-specific webpage supervision, leaving demanding interactive artifact optimization underexplored. Using mini-games as a challenging testbed, we expose limitations in interaction experience, reward signal, and credit assignment fidelity that are obscured by shallow webpage interactions. We introduce WebCatalyst, a browser-interaction reinforcement learning framework with three complementary components: a game-level GUI assessor for sustained interaction and evaluation, Visual Anchor for mitigating VLM reward hacking, and Action-Trace Credit for grounding token-level credit assignment in runtime execution traces. Trained on fewer than 1K requests, predominantly for games, WebCatalyst raises the Cookie-Bench Mini Games score from 68.8 to 75.3 and also the WebGen-Bench accuracy from 48.8 to 56.9 on Qwen3.6-35B-A3B, outperforming Claude Opus 4.7 in average rank across evaluated metrics. Further analysis shows that Visual Anchor prevents aesthetic degradation, while runtime-grounded credit assignment accelerates functional improvement. Together, our results show that improving feedback fidelity is the key to substantially advancing interactive HTML generation.