Factuality on Demand: Controlling the Factuality-Informativeness Trade-off in Text Generation

2026-01-31

AI Summary

This work addresses the inherent trade-off between factual accuracy and informativeness in large language model (LLM) text generation, where different applications impose varying demands on these qualities. To this end, the authors propose a Factuality-Controlled Generation (FCG) framework that, for the first time, enables users to explicitly specify factuality constraints through natural language queries. The approach leverages synthetically generated data to train models capable of dynamically balancing factual correctness with informational richness. The framework introduces a tunable mechanism for factuality control and establishes a comprehensive evaluation protocol that jointly assesses constraint adherence and output informativeness. Experimental results demonstrate that models trained with the proposed synthetic data significantly enhance their ability to satisfy user-defined factuality requirements while preserving high levels of information content.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Large language models (LLMs) encode knowledge with varying degrees of confidence. When responding to queries, models face an inherent trade-off: they can generate responses that are less informative but highly factual, or more informative but potentially less accurate. Different applications demand different balances between informativeness and factuality. We introduce Factuality-Controlled Generation (FCG), a framework that enables users to specify factuality constraints alongside their queries. We propose to evaluate FCG performance on two dimensions: adherence to factuality constraints and response informativeness. We propose to train models on the FCG task using synthetic data, and show that our synthetic training significantly improves models'ability to both respect factuality requirements and maintain informativeness in their outputs.

Problem

• factuality
• informativeness
• text generation
• large language models
• trade-off

Innovation

• Factuality-Controlled Generation
• factuality-informativeness trade-off
• synthetic training data
• controllable text generation
• large language models