[Submitted on 5 Mar 2026 (v1), last revised 6 Nov 2026 (this version, v2)]
Abstract:Large Language Models face a multi-objective optimization problem in constrained long-text generation, balancing global structure, local coherence, and constraint satisfaction. We propose LoopWeaver, a hierarchical workflow with two sequential offline preference-optimization stages. Constraint-aware planning produces structured plans, feasibility filtering removes invalid plans, and a shared deterministic reward constructs preference pairs for plan-level and generation-level DPO. At inference time, LoopWeaver remains feed-forward: it generates and revises a complete plan, filters the revised plan, and then generates segments conditioned on the current sub-plan and running narrative. We evaluate the method on four LongGenBench-style structured scenarios across multiple backbones. Results show improvements in both constraint accuracy and text quality across all four open-source backbones. Our software and data are publicly available at this https URL.
Submission history
From: Guanting Chen [view email]
[v1] Thu, 5 Mar 2026 09:41:39 UTC (3,160 KB)
[v2] Fri, 6 Nov 2026 14:59:04 UTC (5,319 KB)
