World Wires · story 8608 · corroborated · 1 source(s)

WUSH‑KV: KV Cache Quantization with Data‑Adaptive Transforms

WUSH‑KV introduces a low‑bit quantization technique for transformer KV caches using data‑adaptive transforms to reduce memory and bandwidth costs.

Open in the desk

Coverage

• WUSH‑KV: KV Cache Quantization with Data‑Adaptive Transforms arXiv: new preprints (security, AI, physics) · tier C · clear net · 2026-09-29 17:50 UTC

What this site indexes