Rep2Text recovers roughly half the tokens from 16-token sequences using only the last-token representation while retaining semantic coherence across several LLMs.
John D., 56, admitted 04-12-2009 for chest pain
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Rep2Text: Decoding Full Text from a Single LLM Token Representation
Rep2Text recovers roughly half the tokens from 16-token sequences using only the last-token representation while retaining semantic coherence across several LLMs.