LongReD: Mitigating Short-Text Degradation of Long-Context Large Language Models via Restoration Distillation

VLM
Re-assign community