PREM: Prefix-Steered Recurrent Memory for Long-Video Understanding
Long-video understanding must capture transient visual evidence under strict token budgets, yet existing methods compress frames, append memory tokens, or alter internal key-value (KV) caches. We introduce Prefix-Steered Recurrent Memory (PREM), a memory-token-free framework for frozen vision-language models (VLMs). PR...
Si-Ru Zhong, Qiong-Yan Wang, Xiao-Hui Lv et al.
· 0 citations