HiFlow: Hierarchical Feedback-Driven Optimization for Constrained Long-Form Text Generation

2026-03-05

AI Summary

This work addresses the challenge that large language models struggle to simultaneously maintain global structural consistency, local semantic coherence, and adherence to complex constraints when generating long-form text. To this end, the authors propose HiFlow, a two-tier optimization framework comprising a planning layer and a generation layer. HiFlow enables dynamic coordination between structural planning and content generation through constraint-aware outline selection and a closed-loop feedback mechanism. By moving beyond the limitations of static planning and offline supervision, the approach significantly improves both text quality and constraint satisfaction rates. Experimental results demonstrate consistent superiority over existing baselines across multiple backbone language models.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Large language models perform well in short text generation but still struggle with long text generation, particularly under complex constraints. Such tasks involve multiple tightly coupled objectives, including global structural consistency, local semantic coherence, and constraint feasibility, forming a challenging constrained optimization problem. Existing approaches mainly rely on static planning or offline supervision, limiting effective coordination between global and local objectives during generation. To address these challenges, we propose HiFlow, a hierarchical feedback-driven optimization framework for constrained long text generation. HiFlow formulates generation as a two-level optimization process, consisting of a planning layer for global structure and constraint modeling, and a generation layer for conditioned text generation. By incorporating constraint-aware plan screening and closed-loop feedback at both levels, HiFlow enables joint optimization of planning quality and generation behavior, progressively guiding the model toward high-quality, constraint-satisfying outputs. Experiments on multiple backbones confirm HiFlow's effectiveness over baseline methods.

Problem

• constrained long-form text generation
• global structural consistency
• local semantic coherence
• constraint feasibility
• hierarchical optimization

Innovation

• hierarchical optimization
• feedback-driven generation
• constrained text generation
• long-form text
• closed-loop planning