Reproduce “Phase-wise MLLM Tuning for Multi-framework WebUI Code Generation” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Haoran Ma, Linxiao Li, Chenyue Wang, Haochen Sui, Jiechao Gao

The authors introduce a training method for generating runnable frontend code from screenshots across four frameworks. It estimates code-pattern differences and alignment between shared-adapter updates, then uses these signals to organize training and gate sharing. On DesignBench, it reports averages of 0.7950 CLIP and 0.9007 CSR, versus 0.7741 and 0.8712 without gating, using one frozen backbone.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.