Training data changes influence small and large language models' losses similarly, so small proxy models can substitute for large models in data attribution and selection.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Small-to-Large Generalization: Data Influences Models Consistently Across Scale
Training data changes influence small and large language models' losses similarly, so small proxy models can substitute for large models in data attribution and selection.