English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The field of natural language processing has reached breakthroughs with theadvent of transformers. They have remained state-of-the-art since then, andthere also has been much research in analyzing, interpreting, and evaluatingthe attention layers and the underlying embedding space. In addition to theself-attention layers, the feed-forward layers in the transformer are aprominent architectural component. From extensive research, we observe that itsrole is under-explored. We focus on the latent space, known as the ActivationSpace, that consists of the neuron activations from these feed-forward layers.In this survey paper, we review interpretability methods that examine thelearnings that occurred in this activation space. Since there exists onlylimited research in this direction, we conduct a detailed examination of eachwork and point out potential future directions of research. We hope our workprovides a step towards strengthening activation space analysis.

Interpretability in Activation Space Analysis of Transformers: A Focused Survey