Unlocking Anticipatory Text Generation: A Constrained Approach for Large Language Models Decoding

TL;DR

This paper introduces a future-constrained generation framework for large language models to reduce undesired behaviors and improve faithfulness, validated across multiple text generation tasks.

Contribution

It formalizes text generation as a future-constrained problem and uses LLMs to estimate constraint satisfaction, enhancing control over generated content.

Findings

Effective in reducing toxicity and hallucinations

Improves factual accuracy in question-answering

Enhances adherence to input constraints

Abstract

Large Language Models (LLMs) have demonstrated a powerful ability for text generation. However, achieving optimal results with a given prompt or instruction can be challenging, especially for billion-sized models. Additionally, undesired behaviors such as toxicity or hallucinations can manifest. While much larger models (e.g., ChatGPT) may demonstrate strength in mitigating these issues, there is still no guarantee of complete prevention. In this work, we propose formalizing text generation as a future-constrained generation problem to minimize undesirable behaviors and enforce faithfulness to instructions. The estimation of future constraint satisfaction, accomplished using LLMs, guides the text generation process. Our extensive experiments demonstrate the effectiveness of the proposed approach across three distinct text generation tasks: keyword-constrained generation (Lin et al., 2020), toxicity reduction (Gehman et al., 2020), and factual correctness in question-answering (Gao et al., 2023).