The paper introduces GRKV, a training‑free method for compressing the key‑value cache in long‑context large language models. GRKV uses ridge‑regression to redistribute information from evicted tokens to retained ones, aiming to minimize the difference between compressed‑cache and full‑cache attention outputs. Experiments on LongBench and RULER show that GRKV improves overall performance with minimal overhead compared to other merging methods.
arXiv:2609.39329v1 Announce Type: new Abstract: Long-context inference with Large Language Models (LLMs) is bottlenecked by the linearly growing memory of the key-value (KV) cache. Existing compressi...
Long-context large language model (LLM) inference is increasingly constrained by the memory footprint and decoding cost of key-value (KV) caches, limiting sustainable deployment on resource-constrained hardware. Existing KV cache eviction methods typically apply heuristic token scoring over all heads in GQA-based LLMs.
arXiv:2606. 24467v1 Announce Type: new Abstract: Long-context large language model (LLM) inference is increasingly constrained by the memory footprint and decoding cost of key-value (KV) caches, limiting sustainable deployment on resource-constrained hardware.
arXiv:2607. 16213v1 Announce Type: new Abstract: Large Language Models (LLMs) generate text autoregressively, relying on a key-value (KV) cache whose memory footprint grows linearly with context length, creating a major bottleneck.
arXiv:2606. 29563v1 Announce Type: cross Abstract: Large language models (LLMs) excel at complex tasks like question answering and summarization, thanks to their ability to handle long-context inputs.
arXiv:2608.23843v1 Announce Type: new Abstract: Long-context inference in large language models (LLMs) is increasingly limited by the memory required for the key-value (KV) cache. KV cache compressio...
arXiv:2605. 25475v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) are increasingly expected to operate over long contexts, yet standard softmax attention incurs a KV cache that grows linearly with sequence length, quickly becoming the bottleneck for long context inference.
DropKV introduces a new KV‑cache eviction framework that decouples residual‑output perturbation into independent per‑token scores, avoiding the NP‑hard combinatorial optimization of prior methods. The approach offers a provable constant‑factor approximation under a cone condition verified on three long‑context LLMs, and a high‑performance Triton‑based implementation that yields a 19.8× speedup in scoring kernels and up to 9.9% end‑to‑end prefill speedup. Benchmarks on RULER and LongBench show DropKV consistently outperforms state‑of‑the‑art baselines at the same cache budgets, achieving the lowest inference latency among eviction strategies while remaining faster than dynamic cache baselines.
arXiv:2602.08005v2 Announce Type: replace-cross Abstract: Efficient long-context inference faces two coupled bottlenecks: KV-cache memory grows linearly with context length, while attention computati...
NestedKV is a training‑free key‑only KV cache compression technique for long‑context language models that uses global, block‑level, and sliding‑window key anchors to score tokens via multi‑time‑scale cosine anomaly. It combines these rankings with a head‑adaptive outer learner and surprise‑gated token routing, requiring no model modification or additional training. Experiments on Qwen3 and Llama‑3.2 across benchmarks such as RULER, LongBench, and MMLU‑Pro show that NestedKV outperforms existing methods when the retained cache is small, achieving up to 19‑point gains on RULER and LongBench at a 75% retention rate.
arXiv:2605.23258v2 Announce Type: replace Abstract: KV cache growth is a major bottleneck for long-context inference in large language models. Existing methods are often dominated by binary eviction...
DropKV introduces a new key-value cache eviction strategy for large language models that decouples the joint eviction decision into independent per-token scoring, thereby avoiding the NP-hard combinatorial problem of minimizing output perturbation. The method offers a provable constant-factor approximation guarantee under a cone condition verified on three long-context LLMs, and its high-performance implementation using fused Triton kernels achieves a 19.8× speedup in scoring and up to 9.9% end-to-end prefill speedup. Evaluations on RULER and LongBench show that DropKV outperforms state-of-the-art baselines at equivalent cache budgets, delivering the lowest inference latency among eviction methods while remaining faster than the dynamic cache baseline.