C$^3$TG: Conflict-aware, Composite, and Collaborative Controlled Text Generation

2025-11-12

AI Summary

To address attribute conflicts, coupling, and the absence of iterative optimization in multi-attribute text generation with large language models (LLMs), this paper proposes a two-stage fine-grained multidimensional control framework. In the generation stage, selective token-level control is achieved via weighted KL divergence regularization and a 17-dimensional attribute classifier. In the optimization stage, a conflict-aware energy function integrates classifier scores with penalty terms to dynamically reconcile competing attributes and support iterative refinement. The method requires no architectural modifications to the underlying LLM, enabling flexible, plug-and-play multi-attribute coordination. Experiments demonstrate significant improvements over baselines in attribute accuracy, fluency, and lexical diversity, while effectively reducing textual toxicity—validating both the efficacy and robustness of multidimensional controllable generation.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Recent advancements in large language models (LLMs) have demonstrated remarkable text generation capabilities. However, controlling specific attributes of generated text remains challenging without architectural modifications or extensive fine-tuning. Current methods typically toggle a single, basic attribute but struggle with precise multi-attribute control. In scenarios where attribute requirements conflict, existing methods lack coordination mechanisms, causing interference between desired attributes. Furthermore, these methods fail to incorporate iterative optimization processes in the controlled generation pipeline. To address these limitations, we propose Conflict-aware, Composite, and Collaborative Controlled Text Generation (C$^3$TG), a two-phase framework for fine-grained, multi-dimensional text attribute control. During generation, C$^3$TG selectively pairs the LLM with the required attribute classifiers from the 17 available dimensions and employs weighted KL-divergence to adjust token probabilities. The optimization phase then leverages an energy function combining classifier scores and penalty terms to resolve attribute conflicts through iterative feedback, enabling precise control over multiple dimensions simultaneously while preserving natural text flow. Experiments show that C$^3$TG significantly outperforms baselines across multiple metrics including attribute accuracy, linguistic fluency, and output diversity, while simultaneously reducing toxicity. These results establish C$^3$TG as an effective and flexible solution for multi-dimensional text attribute control that requires no costly model modifications.

Problem

• Achieving fine-grained multi-attribute text generation control
• Resolving attribute conflicts during controlled text generation
• Enabling precise multi-dimensional control without model modifications

Innovation

• Two-phase framework for multi-dimensional text control
• Weighted KL-divergence adjusts token probabilities during generation
• Energy function resolves conflicts through iterative optimization