AIbase
Product LibraryTool Navigation

HiRED

Public

[AAAI 2025] HiRED strategically drops visual tokens in the image encoding stage to improve inference efficiency for High-Resolution Vision-Language Models (e.g., LLaVA-Next) under a fixed token budget.

Creat2024-08-20T00:43:46
Update2025-03-19T18:16:45
https://www.arxiv.org/abs/2408.10945
28
Stars
0
Stars Increase

Related projects