ReLook: Vision-Grounded RL with a Multimodal LLM Critic for Agentic Web Coding

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Li, Yuhang, Zhang, Chenchen, Lv, Ruilin, Liu, Ao, Deng, Ken, Zhang, Yuanxing, Liu, Jiaheng, Zhou, Wiggin, Zhou, Bo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914090333503488
author Li, Yuhang
Zhang, Chenchen
Lv, Ruilin
Liu, Ao
Deng, Ken
Zhang, Yuanxing
Liu, Jiaheng
Zhou, Wiggin
Zhou, Bo
author_facet Li, Yuhang
Zhang, Chenchen
Lv, Ruilin
Liu, Ao
Deng, Ken
Zhang, Yuanxing
Liu, Jiaheng
Zhou, Wiggin
Zhou, Bo
contents While Large Language Models (LLMs) excel at algorithmic code generation, they struggle with front-end development, where correctness is judged on rendered pixels and interaction. We present ReLook, an agentic, vision-grounded reinforcement learning framework that empowers an agent to close a robust generate--diagnose--refine loop by invoking a multimodal LLM (MLLM) as a tool. During training, the agent uses the MLLM-in-the-loop both as a visual critic--scoring code with screenshots--and as a source of actionable, vision-grounded feedback; a strict zero-reward rule for invalid renders anchors renderability and prevents reward hacking. To prevent behavioral collapse, we introduce Forced Optimization, a strict acceptance rule that admits only improving revisions, yielding monotonically better trajectories. At inference, we decouple the critic and run a lightweight, critic-free self-edit cycle, keeping latency comparable to base decoding while retaining most of the gains. Across three widely used benchmarks, ReLook consistently outperforms strong baselines in vision-grounded front-end code generation, highlighting the benefits of agentic perception, visual rewards, and training-inference decoupling.
format Preprint
id arxiv_https___arxiv_org_abs_2510_11498
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ReLook: Vision-Grounded RL with a Multimodal LLM Critic for Agentic Web Coding
Li, Yuhang
Zhang, Chenchen
Lv, Ruilin
Liu, Ao
Deng, Ken
Zhang, Yuanxing
Liu, Jiaheng
Zhou, Wiggin
Zhou, Bo
Machine Learning
Computation and Language
While Large Language Models (LLMs) excel at algorithmic code generation, they struggle with front-end development, where correctness is judged on rendered pixels and interaction. We present ReLook, an agentic, vision-grounded reinforcement learning framework that empowers an agent to close a robust generate--diagnose--refine loop by invoking a multimodal LLM (MLLM) as a tool. During training, the agent uses the MLLM-in-the-loop both as a visual critic--scoring code with screenshots--and as a source of actionable, vision-grounded feedback; a strict zero-reward rule for invalid renders anchors renderability and prevents reward hacking. To prevent behavioral collapse, we introduce Forced Optimization, a strict acceptance rule that admits only improving revisions, yielding monotonically better trajectories. At inference, we decouple the critic and run a lightweight, critic-free self-edit cycle, keeping latency comparable to base decoding while retaining most of the gains. Across three widely used benchmarks, ReLook consistently outperforms strong baselines in vision-grounded front-end code generation, highlighting the benefits of agentic perception, visual rewards, and training-inference decoupling.
title ReLook: Vision-Grounded RL with a Multimodal LLM Critic for Agentic Web Coding
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2510.11498