# RoLaQ improves 2-bit KV cache accuracy and speed

> The new RoLaQ method boosts accuracy and decoding throughput for low‑bit key‑value caches in multi‑head latent attention models.

Oossa · 2026-10-16 · https://oossa.com/en/rolaq-improves-2-bit-kv-cache-accuracy-and-speed

Researchers led by Shiyu Ji released RoLaQ, a calibration technique for 2‑bit key‑value caches used in multi‑head latent attention. The method jointly adjusts rotation and quantization range while keeping model weights frozen. In tests on three models and six benchmarks, RoLaQ gave the highest average accuracy among 2‑bit methods. On DeepSeek‑V2‑Lite, an SGLang implementation reached 3.96 × the BF16 cache capacity and 2.51 × the decode speed with a 256K token context.

## The facts

- RoLaQ was submitted on 14 Oct 2026 (arXiv ID 2610.20686).
- It achieved 3.96× KV capacity and 2.51× decode throughput on DeepSeek‑V2‑Lite.

## Why it matters

Developers of long‑context language models can run larger prompts faster without extra hardware.

## Sources & references

1. [RoLaQ: Joint Rotation and Range Calibration for 2-Bit KV Caches in Multi-Head Latent Attention](https://arxiv.org/abs/2610.20686) – arXiv cs.LG, 2026-10-16

Last updated: 2026-10-16