方法对比

	可解释句子嵌入	可解释循环神经网络
领域	深度学习	深度学习
方法族	Machine learning	Machine learning
起源年份	2016–2018	2017–2020
提出者	Conneau et al.; Ribeiro et al. (probing + LIME frameworks)	Arrived via XAI literature (Arrieta et al., Lundberg & Lee, and attention-based RNN work)
类型	Post-hoc interpretability applied to sentence encoders	Interpretability framework applied to sequence models
开创性文献	Conneau, A., Kruszewski, G., Lample, G., Barrault, L., & Baroni, M. (2018). What you can cram into a single $\vec{v}$ector: Probing sentence embeddings for linguistic properties. In Proceedings of ACL 2018, pp. 2126–2136.	Arrieta, A. B., Diaz-Rodriguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., Garcia, S., Gil-Lopez, S., Molina, D., Benjamins, R., Chatila, R., & Herrera, F. (2020). Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI. Information Fusion, 58, 82–115.
别名	interpretable sentence representations, XAI sentence embeddings, probing sentence embeddings, explainable sentence vectors	Explainable RNN, Interpretable RNN, XAI-RNN, Transparent Recurrent Neural Network
相关	6	5
摘要	Explainable sentence embeddings combine dense sentence representation learning with post-hoc or intrinsic interpretability tools — such as probing classifiers, LIME, SHAP, or attention attribution — to reveal what linguistic and semantic information is encoded in a sentence vector and why a downstream model makes a given prediction. The goal is to retain the representational power of modern encoders while making their behavior auditable.	An Explainable Recurrent Neural Network (XAI-RNN) pairs a standard RNN architecture with a post-hoc or intrinsic interpretability method — such as SHAP, LIME, integrated gradients, or attention visualization — to reveal which input time steps or tokens most influence the model's sequential predictions, without sacrificing predictive accuracy.