Seamless Packing, a sliding-window plus first-fit-decreasing packing strategy, improves continual pre-training results over concatenation-truncation and best-fit-decreasing baselines by small margins.
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
CONDITIONAL 1representative citing papers
citing papers explorer
-
Improving Continual Pre-training Through Seamless Data Packing
Seamless Packing, a sliding-window plus first-fit-decreasing packing strategy, improves continual pre-training results over concatenation-truncation and best-fit-decreasing baselines by small margins.