The paper introduces a multi‑shell decoder for Leech‑lattice vector quantization, achieving the best reported 2‑bit quality under its evaluation protocol. It presents a GPU‑friendly layout that fuses dequantization with matrix‑vector multiplication, demonstrating significant speed and memory advantages over traditional one‑hot masks and other 4‑bit methods. Experiments show the new kernel outperforms baseline approaches across multiple model sizes, with measurable gains in throughput and reduced byte traffic.
By Pier-Jean Malandrino (Scub)
arXiv:2608. 07915v1 Announce Type: new Abstract: Large language models (LLMs) increasingly read long inputs in the agentic era, from whole documents and codebases to conversations across many turns.
By Jiamu Zhang, Liang Wu, Kelly Wan, Hanjie Chen, Liangjie Hong
arXiv:2607. 07144v1 Announce Type: new Abstract: The key-value (KV) cache dominates the memory cost of long-context autoregressive inference, and a growing body of work compresses it through quantization, eviction, or offloading.
By Vladimir Gusev
arXiv:2606. 20474v1 Announce Type: cross Abstract: Context-heavy agents place unusual pressure on the key-value (KV) cache: long prefixes are reused across many short turns, while concurrency determines whether the serving system can keep GPUs utilized.
By Inesh Chakrabarti (Advanced Micro Devices, University of California, Los Angeles), David Limpus (Advanced Micro Devices, Purdue University), Aditi Ghai Rana (Advanced Micro Devices), Bowen Bao (Advanced Micro Devices), Spandan Tiwari (Advanced Micro Devices), Thiago Crepaldi (Advanced Micro Devices), Ashish Sirasao (Advanced Micro Devices)
arXiv:2606. 21633v2 Announce Type: replace Abstract: The KV cache dominates GPU memory in long-context LLM serving, crowding out batch capacity and leaving GPU compute idle.
By Omin Kwon, Doyeon Kim, Jongseok Park, Seung Yul Lee, Ion Stoica, Jae W. Lee
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
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)
arXiv:2606. 24033v1 Announce Type: new Abstract: Existing low-bit KV-cache quantizers often treat each cached key as a flat vector.
By Fengfeng Liang, Yuechen Zhang, Jiaya Jia
arXiv:2606. 09916v1 Announce Type: cross Abstract: Multi-turn LLM agents fan short queries into long trajectories of tool calls, search results, and intermediate reasoning.
By Junjie Li, Jiong Lou, Jie Li
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
Serving a 235B-parameter Mixture-of-Experts (MoE) model on a single 8 GB GPU is bottlenecked not by compute but by memory bandwidth: decode must stream each token's active experts from whichever tier holds them, and on consumer hardware most experts sit on an SSD far slower than RAM. We quantify this bandwidth wall on Qwen3-235B (Q4_K_M, 134 GB): measured decode is 0.
The paper investigates how prefix caching, a default optimization in open‑source LLM serving stacks, affects reproducibility when combined with weight quantization. Experiments on an eighty‑episode multi‑turn agentic tool‑use workload show that enabling the cache causes the agent’s trajectory to change in 36.2 % of episodes at 16‑bit precision and 75.0 % at 4‑bit precision, while disabling the cache yields perfectly reproducible runs. The study identifies specific cache‑related settings that drive run‑to‑run divergence and demonstrates that cached serving is deterministic only when the cache state is preserved, which is not the case in typical deployments.
By Aditi Patodiya