SeDi-Instruct generates instruction data by relaxing duplicate filtering, sampling cluster-balanced batches, and replacing low-scoring seeds with instructions from high-gradient-norm batches.
E.; Stoica, I.; and Xing, E
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
SeDi-Instruct: Enhancing Alignment of Language Models through Self-Directed Instruction Generation
SeDi-Instruct generates instruction data by relaxing duplicate filtering, sampling cluster-balanced batches, and replacing low-scoring seeds with instructions from high-gradient-norm batches.