HiFlow: Hierarchical Feedback-Driven Optimization for Constrained Long-Form Text Generation
This work tackles the problem of generating high-quality, long-form text that adheres to complex constraints, which is a significant challenge for large language models and affects applications requiring structured or controlled text output.
This paper addresses the challenge of long-form text generation under complex constraints, which often leads to issues with global consistency, local coherence, and constraint feasibility. The authors propose HiFlow, a hierarchical feedback-driven optimization framework that uses a two-level optimization process with constraint-aware plan screening and closed-loop feedback to jointly optimize planning and generation, leading to high-quality, constraint-satisfying outputs.
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.