arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2312.06149cs.CLcs.AI
收藏

解锁预期文本生成:一种面向大语言模型解码的约束方法

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

Salesforce AI Research(Salesforce AI研究部门)

Lifu Tu, Semih Yavuz, Jin Qu, Jiacheng Xu, Rui Meng, Caiming Xiong, Yingbo Zhou

AI总结:

本文提出将文本生构建模为未来约束生成问题,利用 LLM 评估后续约束满足并指导解码,从而提升指令遵循度、降低毒性与幻觉,并在三类任务上验证有效。

摘要:

大语言模型(LLM)已展现出强大的文本生成能力。然而,对于给定提示或指令,要获得最优结果仍具挑战,尤其是十亿参数规模的模型。此外,还可能出现毒性或幻觉等不良行为。尽管更大的模型(例如 ChatGPT)在缓解这些问题方面表现出优势,但仍无法保证完全避免。在本研究中,我们提出将文本生成形式化为一个未来约束生成问题,以最小化不良行为并强制模型忠实遵循指令。该方法利用 LLM 估计未来约束满足情况,并以此指导文本生成过程。大量实验表明,所提方法在三类不同文本生成任务上均有效:关键词约束生成(Lin et al., 2020)、毒性降低(Gehman et al., 2020)以及问答中的事实正确性(Gao et al., 2023)。

英文摘要:

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).

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑