From Sub-Ability Diagnosis to Human-Aligned Generation: Bridging the Gap for Text Length Control via MARKERGEN

TL;DR

This paper introduces MarkerGen, a novel approach that improves length-controlled text generation in large language models by decomposing sub-abilities, explicitly modeling length, and employing a three-stage generation scheme, leading to better adherence and quality.

Contribution

It presents a bottom-up decomposition of length control sub-abilities and a plug-and-play MarkerGen method that enhances LCTG performance and generalizability.

Findings

Significant improvement in length adherence across various settings.

Effective external tool integration enhances LLM capabilities.

Three-stage generation scheme maintains content quality while controlling length.

Abstract

Despite the rapid progress of large language models (LLMs), their length-controllable text generation (LCTG) ability remains below expectations, posing a major limitation for practical applications. Existing methods mainly focus on end-to-end training to reinforce adherence to length constraints. However, the lack of decomposition and targeted enhancement of LCTG sub-abilities restricts further progress. To bridge this gap, we conduct a bottom-up decomposition of LCTG sub-abilities with human patterns as reference and perform a detailed error analysis. On this basis, we propose MarkerGen, a simple-yet-effective plug-and-play approach that:(1) mitigates LLM fundamental deficiencies via external tool integration;(2) conducts explicit length modeling with dynamically inserted markers;(3) employs a three-stage generation scheme to better align length constraints while maintaining content quality. Comprehensive experiments demonstrate that MarkerGen significantly improves LCTG across various settings, exhibiting outstanding effectiveness and generalizability.