paper-with-me

Papers

KVpop -- Key-Value Cache Compression with Predictive Online Pruning

2026-07-06 · Lukas Hauzenberger, Niklas Schmidinger, Anamaria-Roberta Hartl, David Stap, Thomas Schmied, Sebastian Böck, Günter Klambauer, Sepp Hochreiter arxiv

Key-value (KV) cache growth is a major bottleneck in autoregressive decoding, as memory and bandwidth scale linearly with context length. Existing KV eviction methods often rely on static heuristics or proxy scores, which poorly track future token utility and cause brittle eviction as relevance shifts. To address this, we introduce KVpop, which learns a fixed-budget KV eviction policy by directly supervising the keep-or-drop decision. The scorer is trained against a novel future-attention target, computed efficiently without materializing dense attention maps. We further introduce a delayed memory-based scorer that, uniquely among learned eviction methods, defers scoring for a fixed number of steps to exploit near-future context. On AIME and HMMT mathematical reasoning, KVpop retains 98% of full-attention performance on Qwen3-4B at 75% KV cache compression and 97% at 88% compression, consistently outperforming established eviction baselines. Qwen3-8B shows even stronger results, reaching near-full teacher performance. These results show that supervising eviction with future-attention signals cuts memory costs while maintaining quality.

📄 PDF Abstract BibTeX arXiv:2607.05061

Code (0)

등록된 구현이 없습니다.

Tasks

Mathematical Reasoning

Similar Papers 제목 키워드 기반

Information-Aware KV Cache Compression for Long Reasoning

2026-06-25 · Jushi Kai, Zhuiri Xiao, Alexandra Birch, Zhouhan Lin arxiv

Reasoning capability has advanced rapidly in large language models (LLMs), leading to an increasing size of key-value (KV) cache in both prefilling and decoding stages. Existing KV cache compression methods mainly rely o…

Dynamic Memory Compression: Retrofitting LLMs for Accelerated Inference

2024-03-14 · Piotr Nawrot, Adrian Łańcucki, Marcin Chochowski, David Tarjan 외

Transformers have emerged as the backbone of large language models (LLMs). However, generation remains inefficient due to the need to store in memory a cache of key-value representations for past tokens, whose size scale…

GPU

MuKV: Multi-Grained KV Cache Compression for Long Streaming Video Question-Answering

2026-05-21 · Junbin Xiao, Jiajun Chen, Tianxiang Sun, Xun Yang 외 arxiv

Long streaming video QA remains challenging due to growing visual tokens and limited reasoning length of large language models (LLMs). KV-caching stores the Key-Value (KV) of the historical tokens via LLM prefill and ena…

Answer Generation

Value-Guided KV Compression for LLMs via Approximated CUR Decomposition

2025-09-18 · Ayan Sengupta, Siddhant Chaudhary, Tanmoy Chakraborty arxiv

Key-value (KV) cache compression has emerged as a critical technique for reducing the memory and latency overhead of autoregressive language models during inference. Prior approaches predominantly rely on query-key atten…

DepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache Compression

2026-07-07 · Anna Cordoba, Adam Puente Tercero, Nerea Angulo Hijo, Mar Linares Tercero 외 arxiv

Long-context language model inference is increasingly limited by the memory bandwidth and capacity required to store key-value caches, yet existing compression methods often apply uniform budgets across layers or tokens …