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 2406.01468cs.CLcs.AIcs.LG
收藏

理解输出嵌入中的Token概率编码

Understanding Token Probability Encoding in Output Embeddings

Japan Advanced Institute of Science and Technology(日本先进科学技术学院);RIKEN(理化学研究所)

Hakaze Cho, Yoshihiro Sakai, Kenshiro Tanaka, Mariko Kato, Naoya Inoue

AI总结:

该研究发现语言模型输出嵌入中存在准确且稀疏的token概率对数线性编码,可通过调控该编码修改输出分布,且超30%的输出嵌入维度可删除而不显著影响生成效果,还发现预训练早期输出嵌入就已捕获语料token频率信息。

摘要:

本文研究了语言模型输出嵌入中的输出token概率信息。我们发现输出嵌入向量中存在一种近似通用的对数线性编码方式来表征输出token概率,并通过实验证明该编码准确且具有稀疏性。作为因果性验证,我们调控输出嵌入中的编码,能够精准修改输出概率分布。此外,我们在输出概率编码中发现的稀疏性表明,输出嵌入中大量维度对因果语言建模没有贡献。因此,我们尝试删除与输出无关的维度,发现可删除超过30%的维度,且不会导致输出分布和序列生成出现显著偏移。另外,在语言模型的预训练动态过程中,我们发现输出嵌入在训练早期阶段、甚至在参数开始明显收敛之前,就已经捕获了语料库的token频率信息。

英文摘要:

In this paper, we investigate the output token probability information in the output embedding of language models. We find an approximate common log-linear encoding of output token probabilities within the output embedding vectors and empirically demonstrate that it is accurate and sparse. As a causality examination, we steer the encoding in output embedding to modify the output probability distribution accurately. Moreover, the sparsity we find in output probability encoding suggests that a large number of dimensions in the output embedding do not contribute to causal language modeling. Therefore, we attempt to delete the output-unrelated dimensions and find more than 30% of the dimensions can be deleted without significant movement in output distribution and sequence generation. Additionally, in the pre-training dynamics of language models, we find that the output embeddings capture the corpus token frequency information in early steps, even before an obvious convergence of parameters starts.

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。

↑