论文标题
影响解释伯特信息流的模式
Influence Patterns for Explaining Information Flow in BERT
论文作者
论文摘要
尽管您可能需要的全部关注可能证明是正确的，但我们不知道为什么：基于注意力的变压器模型（例如BERT）是优越的，但是信息从输入令牌转向输出预测的流动方式尚不清楚。我们引入了影响模式的影响模式，通过变压器模型的一组路径集。模式量化并将信息流定位到通过一系列模型节点的路径。在实验上，我们发现BERT中的大量信息流通过跳过连接而不是注意力头。我们进一步表明，跨实例的模式的一致性是伯特表现的指标。最后，我们证明模式比以前的基于注意力的方法和基于层的方法的模型性能要大得多。
While attention is all you need may be proving true, we do not know why: attention-based transformer models such as BERT are superior but how information flows from input tokens to output predictions are unclear. We introduce influence patterns, abstractions of sets of paths through a transformer model. Patterns quantify and localize the flow of information to paths passing through a sequence of model nodes. Experimentally, we find that significant portion of information flow in BERT goes through skip connections instead of attention heads. We further show that consistency of patterns across instances is an indicator of BERT's performance. Finally, We demonstrate that patterns account for far more model performance than previous attention-based and layer-based methods.
