Testing Pre-trained Language Models' Understanding of Distributivity via Causal Mediation Analysis
This work addresses the interpretability of language models for researchers, but it is incremental as it focuses on a specific semantic phenomenon.
The paper tackled the problem of assessing pre-trained language models' understanding of distributivity by introducing DistNLI, a diagnostic dataset, and using causal mediation analysis to quantify model behavior, finding that understanding correlates with model and vocabulary size.
To what extent do pre-trained language models grasp semantic knowledge regarding the phenomenon of distributivity? In this paper, we introduce DistNLI, a new diagnostic dataset for natural language inference that targets the semantic difference arising from distributivity, and employ the causal mediation analysis framework to quantify the model behavior and explore the underlying mechanism in this semantically-related task. We find that the extent of models' understanding is associated with model size and vocabulary size. We also provide insights into how models encode such high-level semantic knowledge.