arXiv:2610.03027v1 Announce Type: cross Abstract: The key-value (KV) cache becomes a major memory bottleneck in long-context LLM inference, placing substantial pressure on memory capacity and bandwid...
arXiv:2607. 01065v1 Announce Type: new Abstract: The deployment of Large Language Models (LLMs) with extended context windows is increasingly constrained by the linear growth of Key-Value (KV) cache memory.
arXiv:2610.15384v1 Announce Type: cross Abstract: Compressing the key-value (KV) cache is essential for serving long-context large language models (LLMs) under memory constraints. While current scala...
ResidualQuant introduces a 2‑bit residual quantization scheme for KV caches in looped Transformers, leveraging the similarity of KV states across loops. By using the final‑loop KV states as a reference and applying least‑square scaling, rotations, and loop‑wise mixed precision, the method achieves accurate quantization down to INT2. Experiments on various looped Transformer models show that ResidualQuant improves the accuracy‑memory tradeoff, retaining near BF16 accuracy while reducing KV storage by 80.7% and boosting decode throughput and batch size on an RTX 5090.
arXiv:2608.21362v1 Announce Type: new Abstract: Transformer-based large language models (LLMs) incur high prefill latency because key-value (KV) tensors must be recomputed for each request. Existing...
FlashLoop is a training‑free inference framework for Looped Transformers that reduces cross‑loop redundancy by employing token‑sparse updates, sparse attention, and KV‑residual quantization. It exploits observations that, as loops progress, state changes concentrate on a small token subset, attention differences are dominated by a sparse key subset, and KV residuals become amenable to low‑bit quantization. The method achieves lossless accuracy with up to 1.64× speedup and 6× KV‑cache memory reduction across several Looped Transformer models.
MultiAlloc is a profiling‑guided framework that allocates limited quantization resources between layer‑wise precision enhancement and token‑wise preservation for KV cache quantization in large language models. By using lightweight offline profiling to assess layer sensitivity and token importance, it automatically derives an effective precision allocation strategy that satisfies a fixed memory budget without online search. Experiments on reasoning and long‑context benchmarks show that MultiAlloc consistently outperforms existing layer‑wise, token‑wise, and multi‑granularity quantization methods across models such as Mistral, Llama, DeepSeek‑Qwen, and QwQ.
arXiv:2608. 04074v1 Announce Type: cross Abstract: Long-context LLM decoding reads the key-value (KV) cache at every step.
Dual-QK introduces paired non‑orthogonal transforms for queries and keys to enable sharp queries and flat keys, improving 2‑bit quantization and dynamic channel pruning of KV caches. The method balances key scales for INT2 quantization while concentrating query energy for pruning, and supports channel‑0 protection and bucket‑relative RoPE for long‑context accuracy. Experiments on four models across five generative benchmarks show that Dual‑QK outperforms OSCAR at 40% query‑channel sparsity, achieving 6.8× KV‑cache compression, an 8.3× reduction in KV read volume, and up to 3.75× decoding throughput over unpruned BF16.
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....
arXiv:2609.05764v1 Announce Type: cross Abstract: The key-value (KV) cache is the dominant memory bottleneck in long-context large language model (LLM) decoding: every step reads it entirely, so deco...
The paper introduces D-Quant, a KV cache quantization framework that addresses the memory bottleneck of large language models by using a drift mechanism to convert entropy-coded representations into fixed-size bitstreams. This approach leverages the non-uniform distribution of KV cache values—after rotation and normalization, they approximate a normal distribution—allowing entropy coding to assign shorter codewords to frequent symbols while maintaining regular memory layouts suitable for parallel attention kernels. D-Quant thus aims to reduce memory footprint and bandwidth usage without sacrificing performance.