Token-level intrinsic dimension of internal representations correlates with next-token cross-entropy loss across layers in three LLMs; higher-loss prompts live in higher-dimensional token manifolds.
The emergence of clusters in self-attention dynamics,
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.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
The Geometry of Tokens in Internal Representations of Large Language Models
Token-level intrinsic dimension of internal representations correlates with next-token cross-entropy loss across layers in three LLMs; higher-loss prompts live in higher-dimensional token manifolds.