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

TL;DR

HiFlow introduces a hierarchical feedback-driven framework that improves constrained long-form text generation by jointly optimizing planning and generation through a two-level feedback loop.

Contribution

It presents a novel hierarchical optimization approach with feedback mechanisms to better coordinate global planning and local generation under constraints.

Findings

Outperforms baseline methods on multiple benchmarks.

Effectively balances global structure and local coherence.

Enhances constraint satisfaction in long text generation.

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.