inference/research
verifiedhigh · 0.88updated 2026-10-03

PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funneling (Fan et al.)

paper

Fields

id
sources/qlab-pyramidkv-paper
URL
https://arxiv.org/abs/2406.02069
Publisher
arXiv
Kind
paper
Published
2024-06-04
Accessed
2026-10-03
notes
v4. Layer-dependent KV budget; eviction/compression, not quantization.

Notes

v4. Layer-dependent KV budget; eviction/compression, not quantization.

Cross-references

This record references no other records.