On the Transformation of Latent Space in Fine-Tuned NLP Models
This work addresses the interpretability of fine-tuning for NLP researchers, but it is incremental as it builds on existing analysis methods.
The paper tackled the problem of understanding how latent spaces evolve during fine-tuning in NLP models, finding that higher layers shift toward task-specific concepts while lower layers retain generic ones, with some concepts gaining polarity toward output classes and being usable for adversarial triggers.
We study the evolution of latent space in fine-tuned NLP models. Different from the commonly used probing-framework, we opt for an unsupervised method to analyze representations. More specifically, we discover latent concepts in the representational space using hierarchical clustering. We then use an alignment function to gauge the similarity between the latent space of a pre-trained model and its fine-tuned version. We use traditional linguistic concepts to facilitate our understanding and also study how the model space transforms towards task-specific information. We perform a thorough analysis, comparing pre-trained and fine-tuned models across three models and three downstream tasks. The notable findings of our work are: i) the latent space of the higher layers evolve towards task-specific concepts, ii) whereas the lower layers retain generic concepts acquired in the pre-trained model, iii) we discovered that some concepts in the higher layers acquire polarity towards the output class, and iv) that these concepts can be used for generating adversarial triggers.