A controlled study shows RETRO-style models need a minimum input-neighbor overlap to activate, and paraphrased synthetic context can speed up training by about 40% at a small perplexity cost.
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
-
Studying the Role of Input-Neighbor Overlap in Retrieval-Augmented Language Models Training Efficiency
A controlled study shows RETRO-style models need a minimum input-neighbor overlap to activate, and paraphrased synthetic context can speed up training by about 40% at a small perplexity cost.