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.
By Seifeldin Abdellatif
arXiv:2609.26333v1 Announce Type: new
Abstract: Prefill and decode reward different approaches to quantization: low-precision arithmetic accelerates prompt processing, while compact weights reduce me...
By Andrei Panferov, Maximilian Kleinegger, Sweta Priyadarshi, Tijmen Blankevoort, Dan Alistarh
arXiv:2608.30427v1 Announce Type: cross
Abstract: Speculative decoding speeds up generation with an efficient draft model (drafter) that proposes tokens for a target model to verify in one pass, pres...
By Ephrem Wu
The paper investigates whether training Mixture-of-Experts (MoE) routers can improve memory‑bandwidth locality on consumer GPUs. Using a new zero‑surgery telemetry tool, the authors measure that a large Qwen3‑235B model is bottlenecked by disk‑based expert access, and that an LRU cache can serve a majority of requests. They pre‑register experiments training 137 M‑parameter MoE models with locality‑aware losses, finding that while cache misses can drop up to 60 % (99 % static‑pin hit rate), every configuration fails to meet a strict 1 % perplexity threshold, indicating a tight coupling between cache efficiency and model quality.
By Shriniwas Ramesh Suram
arXiv:2609.17109v1 Announce Type: new
Abstract: A common small-model deployment runs one shared backbone with several LoRA specialists that answer over the same context. Serving them naively re-prefi...
By Dushyant Rajput
VestigeKV is a new KV‑cache technique that uses a 64‑dimensional vestigial branch—originally a RoPE component repurposed during NoPE training—as a query‑independent eviction signal. By reading only 11 % of each cache row, the method partitions the cache into an attended tier (top‑m rows) and an archive tier (all other rows), which is GPU‑resident and never deleted. The approach achieves near‑perfect retrieval (1.00 at 8×, 0.92 at 32×) without any training, quantization, or changes to weights or kernels.
By WenJie Fan
arXiv:2608.28911v1 Announce Type: new
Abstract: The key-value (KV) cache is the dominant memory bottleneck of long-context large language model (LLM) inference, growing linearly with context length....
By Daeha Lee, Do-Hyung Kim, Jae-Hong Kim
A common small-model deployment runs one shared backbone with several LoRA specialists that answer over the same context. Serving them naively re-prefills that shared context once per specialist. We s...
arXiv:2607. 28699v1 Announce Type: cross Abstract: KV-cache quantization is validated today by offline benchmark averages; a deployed system cannot tell whether compression is damaging the request it is serving right now.
By Fanzhe Wei, Li Liu
The paper presents a Pareto atlas of LLM inference optimizations, mapping cost, quality, and latency trade‑offs for Qwen2.5‑7B‑Instruct on L4, A100, and H100 GPUs. Using 54 measured configurations and a calibrated simulator, it identifies 18 of 36 setups on the Pareto frontier, showing that combined methods outperform single ones. Quality tests reveal that AWQ 4bit and FP8 weights offer significant latency reductions while largely preserving accuracy, but naive FP8 KV caching fails to answer any questions correctly.
By Srikanta Datta Tumkur, Jay Iyer, Mehar Simhadri, Sai Pavan Kumar, Sai Kapil Kumar, Ramesh Nampelly
arXiv:2609.03949v2 Announce Type: replace-cross
Abstract: A long-lived KV cache must be compressed before the queries that will read it exist. Selection by observed attention collapses there: on a No...
By WenJie Fan
arXiv:2607. 02043v1 Announce Type: cross Abstract: Disaggregated LLM serving runs prefill and decode on separate GPU pools to keep the two phases from interfering.
By Shrikara Arun, Anjaly Parayil, Srikant Bharadwaj, Renee St. Amant, Victor R\"uhle