Training LLMs with injected errors followed by correction tags aims to teach the model to self-correct at inference, demonstrated on a single arithmetic conversion example.
Efficient estima tion of word representations in vector space,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
other 1
citation-polarity summary
fields
cs.AI 1years
2024 1verdicts
REJECT 1roles
other 1polarities
unclear 1representative citing papers
citing papers explorer
-
Internalized Self-Correction for Large Language Models
Training LLMs with injected errors followed by correction tags aims to teach the model to self-correct at inference, demonstrated on a single arithmetic conversion example.