Research map / Multi-agent systems and games / LLM agents
MulFCoder: Framework-conditioned Multi-agent for MLLM-based Multi-framework Front-end Code Generation
Jie Wu, Haoran Ma, Shisong Tang, Yulin Xu, Xiaoyu Kang, Jiechao Gao
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes MulFCoder, a framework-conditioned multi-agent MLLM approach for generating runnable multi-framework (React/Vue/Angular) front-end code from UI screenshots.
Problem
Generating runnable front-end code from UI screenshots has mostly targeted HTML/CSS; multi-framework generation for React, Vue, and Angular is underexplored, and naive prompt changes leave large performance gaps and framework-specific error modes.
Approach
A framework-conditioned multi-agent design for MLLM-based screenshot-to-code generation that addresses framework-specific errors, rather than relying on prompt modification alone. Details beyond the abstract are not available.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add MulFCoder framework-conditioned multi-agent pipeline | GPT-4o · Design2Code Angular | CSR: 0.81 -> 0.90 | Improved |
| add MulFCoder | Qwen2.5-VL-7B · DesignBench Angular | CSR: 0.62 -> 0.75 | Improved |
| add MulFCoder | Qwen2.5-VL-7B · Design2Code Vue | CSR: 0.83 -> 0.91 | Improved |
| add MulFCoder | GPT-4o · DesignBench Angular | CLIP score: 0.76 -> 0.76 | No clear effect |
| add MulFCoder | GPT-4o · DesignBench React | CSR: 0.94 -> 0.98 | Improved |
| add MulFCoder on weaker backbones versus stronger ones | LLaVA-v1.5-7B vs Gemini-2.5-Pro · Design2Code Angular | CSR gain: 0.65 -> 0.78 (LLaVA) vs 0.85 -> 0.91 (Gemini) | Improved |
| add MulFCoder on Vanilla HTML | GPT-4o · Design2Code Vanilla | CLIP score: 0.91 -> 0.93 | Improved |
| apply MulFCoder bounded PATCH repair to Angular failures | MulFCoder · DesignBench Angular | error fix ratio: 31.2% average vs 50.4% React | Hurt |
| raise task difficulty from Easy to Hard | MulFCoder · DesignBench Angular | CSR: 0.95 -> 0.62 (Gemini), 0.81 -> 0.10 (LLaVA) | Hurt |
Closest papers by meaning
- FullStack-Agent: Enhancing Agentic Full-Stack Web Coding via Development-Oriented Testing and Repository Back-Translation (ICML 2026 · similarity 0.84)
- GitTaskBench: A Benchmark for Code Agents Solving Real-World Tasks Through Code Repository Leveraging (AAAI 2026 · similarity 0.82)
- WebGen-Bench: Evaluating LLMs on Generating Interactive and Functional Websites from Scratch (NeurIPS 2025 · similarity 0.82)
- WebGen-Agent: Enhancing Interactive Website Generation with Multi-Level Feedback and Step-Level Reinforcement Learning (ICLR 2026 · similarity 0.82)
- Automated Benchmark Generation for Repository-Level Coding Tasks (ICML 2025 · similarity 0.82)
- Vision2Web: A Hierarchical Benchmark for Visual Website Development with Agent Verification (ICML 2026 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
