Neural-embedding-based data selection underperforms DSIR on most GLUE tasks, and the paper's hybrid method is untested because only α=0 was run.
Title resolution pending
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
-
Investigating the Impact of Data Selection Strategies on Language Model Performance
Neural-embedding-based data selection underperforms DSIR on most GLUE tasks, and the paper's hybrid method is untested because only α=0 was run.