arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2404.00752cs.CLcs.AI
收藏

关于最小贝叶斯风险解码的真实分布逼近研究

On the True Distribution Approximation of Minimum Bayes-Risk Decoding

Nagoya University(名古屋大学);CyberAgent(CyberAgent公司)

Atsumoto Ohashi, Ukyo Honda, Tetsuro Morimura, Yuu Jinnai

AI总结:

本文研究最小贝叶斯风险解码中采样分布逼近真实参考分布的问题,提出用异常检测衡量逼近程度,并首次实证证明异常分数与解码性能变化密切相关。

摘要:

最小贝叶斯风险(MBR)解码近年来在文本生成领域重新受到关注。MBR解码将从模型中采样的文本视为伪参考,并选择与其他文本相似度最高的文本。因此,采样是MBR解码的关键要素之一,此前研究指出性能会因采样方法的不同而变化。从理论角度看,这种性能变化很可能与样本在多大程度上逼近参考的真实分布有关。然而,这一逼近问题尚未得到深入研究。在本研究中,我们提出使用异常检测来衡量逼近程度。我们首先仔细考察了性能变化,然后表明此前关于样本的假设与性能变化的相关性不强,而我们引入的异常分数则与之密切相关。该结果首次从实证上支持了性能与MBR解码核心假设之间的联系。

英文摘要:

Minimum Bayes-risk (MBR) decoding has recently gained renewed attention in text generation. MBR decoding considers texts sampled from a model as pseudo-references and selects the text with the highest similarity to the others. Therefore, sampling is one of the key elements of MBR decoding, and previous studies reported that the performance varies by sampling methods. From a theoretical standpoint, this performance variation is likely tied to how closely the samples approximate the true distribution of references. However, this approximation has not been the subject of in-depth study. In this study, we propose using anomaly detection to measure the degree of approximation. We first closely examine the performance variation and then show that previous hypotheses about samples do not correlate well with the variation, but our introduced anomaly scores do. The results are the first to empirically support the link between the performance and the core assumption of MBR decoding.

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。

↑