Hidden layer distillation yields systematic perplexity gains over logit KD in LLM pre-training but does not consistently improve downstream performance.
Findings of the Association for Computational Linguistics
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Study on Hidden Layer Distillation for Large Language Model Pre-Training
Hidden layer distillation yields systematic perplexity gains over logit KD in LLM pre-training but does not consistently improve downstream performance.