arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2109.14723cs.CL
收藏

BeliefBank:为系统信念概念的预训练语言模型添加记忆

BeliefBank: Adding Memory to a Pre-Trained Language Model for a Systematic Notion of Belief

Allen Institute for AI(艾伦人工智能研究所);Center for Information and Language Processing, LMU Munich(慕尼黑大学信息与语言处理中心)

Nora Kassner, Oyvind Tafjord, Hinrich Schütze, Peter Clark

摘要:

尽管预训练的语言模型(PTLM)包含大量的世界知识,但即使经过专门的训练,它们在探测时仍然会产生不一致的问题答案。因此,很难确定模型实际上“相信”这个世界,这使得它容易受到不一致行为和简单错误的影响。我们的目标是减少这些问题。我们的方法是将PTLM嵌入到一个更广泛的系统中,该系统还包括一个不断发展的、象征性的信念记忆--一个信念银行--它记录但随后可能修改原始PTLM答案。我们描述了两种机制,以提高整个系统的信念一致性。首先,一个推理组件--加权MaxSAT求解器--修改与其他人严重冲突的信念。第二,反馈组件使用已知的信念作为上下文向PTLM发出未来的查询。我们发现,在一个受控的实验环境中,这两种机制导致整个系统中更一致的信念,随着时间的推移,提高了答案的准确性和一致性。这是非常重要的,因为它是迈向基于PTLM的架构的第一步,具有系统的信念概念,使他们能够构建一个更连贯的世界图景,并随着时间的推移而改进,而无需模型再训练。

英文摘要:

Although pretrained language models (PTLMs) contain significant amounts of world knowledge, they can still produce inconsistent answers to questions when probed, even after specialized training. As a result, it can be hard to identify what the model actually "believes" about the world, making it susceptible to inconsistent behavior and simple errors. Our goal is to reduce these problems. Our approach is to embed a PTLM in a broader system that also includes an evolving, symbolic memory of beliefs -- a BeliefBank -- that records but then may modify the raw PTLM answers. We describe two mechanisms to improve belief consistency in the overall system. First, a reasoning component -- a weighted MaxSAT solver -- revises beliefs that significantly clash with others. Second, a feedback component issues future queries to the PTLM using known beliefs as context. We show that, in a controlled experimental setting, these two mechanisms result in more consistent beliefs in the overall system, improving both the accuracy and consistency of its answers over time. This is significant as it is a first step towards PTLM-based architectures with a systematic notion of belief, enabling them to construct a more coherent picture of the world, and improve over time without model retraining.

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。

↑