SFR applies conditional flow matching on future sentence embeddings as a training regularizer to increase output diversity in style-conditioned LLMs without deployment overhead.
Elias Frantar, Saleh Ashkboos, Torsten Hoefler, and Dan Alistarh
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
-
Semantic Flow Regularization: Teaching LLMs to Generate Diverse Yet Coherent Responses
SFR applies conditional flow matching on future sentence embeddings as a training regularizer to increase output diversity in style-conditioned LLMs without deployment overhead.