Uni-Layout: Integrating Human Feedback in Unified Layout Generation and Evaluation

2025-08-04

AI Summary

Existing layout generation approaches suffer from two key limitations: (1) task fragmentation across diverse layout generation scenarios, and (2) misalignment between conventional evaluation metrics and human perceptual preferences. To address these, this paper proposes a unified generation-and-evaluation framework. First, we introduce Layout-HF100k—the first large-scale human feedback dataset for layout generation, containing 100K preference-labeled layout pairs. Second, we propose Dynamic Margin Preference Optimization (DMPO), a novel training paradigm that jointly optimizes the generator and a human-aligned evaluator via preference learning with adaptive margins. Third, we unify multi-task layout generation through natural language prompting, integrating vision-geometric priors, chain-of-thought reasoning, and confidence estimation. Experiments demonstrate significant improvements over both task-specific and general-purpose baselines in both generation quality and evaluation fidelity, with stronger alignment to human preferences. The code is publicly available.

Technology Category

• Humans and AI: Learning Human Values and Preferences
• Natural Language Processing: Generation
• Machine Learning: Learning Preferences or Rankings

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Layout generation plays a crucial role in enhancing both user experience and design efficiency. However, current approaches suffer from task-specific generation capabilities and perceptually misaligned evaluation metrics, leading to limited applicability and ineffective measurement. In this paper, we propose extit{Uni-Layout}, a novel framework that achieves unified generation, human-mimicking evaluation and alignment between the two. For universal generation, we incorporate various layout tasks into a single taxonomy and develop a unified generator that handles background or element contents constrained tasks via natural language prompts. To introduce human feedback for the effective evaluation of layouts, we build extit{Layout-HF100k}, the first large-scale human feedback dataset with 100,000 expertly annotated layouts. Based on extit{Layout-HF100k}, we introduce a human-mimicking evaluator that integrates visual and geometric information, employing a Chain-of-Thought mechanism to conduct qualitative assessments alongside a confidence estimation module to yield quantitative measurements. For better alignment between the generator and the evaluator, we integrate them into a cohesive system by adopting Dynamic-Margin Preference Optimization (DMPO), which dynamically adjusts margins based on preference strength to better align with human judgments. Extensive experiments show that extit{Uni-Layout} significantly outperforms both task-specific and general-purpose methods. Our code is publicly available at https://github.com/JD-GenX/Uni-Layout.

Problem

• Unifies diverse layout generation tasks into a single framework
• Addresses misaligned evaluation metrics with human feedback integration
• Aligns generator and evaluator via dynamic preference optimization

Innovation

• Unified generator via natural language prompts
• Human feedback dataset Layout-HF100k
• Dynamic-Margin Preference Optimization alignment