Unlocking Anticipatory Text Generation: A Constrained Approach for Large Language Models Decoding
This work addresses the challenge of controlling text generation in large language models to reduce toxicity and improve faithfulness, which is important for developers and users of AI systems, though it appears incremental as it builds on existing constraint-based methods.
The paper tackles the problem of undesirable behaviors like toxicity and hallucinations in large language models by formalizing text generation as a future-constrained generation problem, using LLMs to estimate constraint satisfaction and guide generation, with experiments showing effectiveness across tasks such as keyword-constrained generation, toxicity reduction, and factual correctness in question-answering.
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).