A proof and empirical check that cosine similarity in language model representations encodes the intrinsic geometry of features through shortest paths on manifolds.
Superintelligence: Paths , Dangers , Strategies
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
The Origins of Representation Manifolds in Large Language Models
A proof and empirical check that cosine similarity in language model representations encodes the intrinsic geometry of features through shortest paths on manifolds.