Top-K attention heads with the highest cosine similarity to a concept vector form a sparse module whose scalar scaling can strengthen or suppress that concept in LLMs and vision transformers.
Explainable ai: A review of machine learning interpretability methods
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
From Concepts to Components: Concept-Agnostic Attention Module Discovery in Transformers
Top-K attention heads with the highest cosine similarity to a concept vector form a sparse module whose scalar scaling can strengthen or suppress that concept in LLMs and vision transformers.