LongReD reduces short-text performance loss after long-context extension by training the extended model to match the original model's hidden states on short texts and using skipped position indices to bridge short and long inputs.
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
-
LongReD: Mitigating Short-Text Degradation of Long-Context Large Language Models via Restoration Distillation
LongReD reduces short-text performance loss after long-context extension by training the extended model to match the original model's hidden states on short texts and using skipped position indices to bridge short and long inputs.