arXiv Computation and Language By Seifeldin Abdellatif

Quality Recovery for Quantized KV Caches via Low-Rank Attention Adaptation

Read the original on arXiv Computation and Language →

The paper investigates how to recover language model quality lost when using low‑bit key–value caches for autoregressive decoding. By keeping the quantizer fixed and distilling the full‑precision cache behavior into low‑rank Q/K/V projection updates, the authors demonstrate that 4‑bit affine‑cache adapters recover roughly 54 % of the perplexity gap on TinyLlama‑1.1B and 76 % on Gemma‑4‑12B, while preserving most long‑context retrieval. Even a 2‑bit rank–token sweep can dramatically reduce TinyLlama’s perplexity, though it only partially restores retrieval performance.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Computation and Language.

arXiv AI
Aug 26

Minima-KV: Retention-Preserving KV Cache Compression with Mixed-Format Paged Attention

Minima-KV introduces a retention‑preserving hierarchy for mixed‑format paged attention that keeps recent and protected anchor pages in FP8 while older pages are compressed into packed TQ3, allowing every live‑request page to remain addressable. The approach uses format‑specific kernels and a globally normalized online‑softmax merge to compute partial attention states, enabling direct heterogeneous decoding without a dense shadow cache. Experiments on Qwen3.6‑27B on a 96‑GB NVIDIA RTX PRO 6000 Blackwell GPU show 3.50× compression over BF16 and 1.75× over FP8, with minimal impact on performance across long‑context benchmarks.

By Sergii Kozyrev (Minima AI, Inc), Davyd Maiboroda (Minima AI, Inc)