How ResidualQuant works: share a 4-bit anchor, store each loop's KV residual in 2 bits

sehoonkim418 · x · 2026-10-12

Related event: KAIST's ResidualQuant: 2-bit Residual Quantization Cuts Looped Transformer KV Cache by 80%(10 posts)→

Original post →

More from Infra

Infra channel →