The paper presents a unified framework for computing surprisal over arbitrary linguistic units, disentangling unit definition from tokenization choices in language models.
Roger Levy
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
On the Proper Treatment of Units in Surprisal Theory
The paper presents a unified framework for computing surprisal over arbitrary linguistic units, disentangling unit definition from tokenization choices in language models.