NeurIPS 2025

R-KV: Redundancy-aware KV Cache Compression for Reasoning Models

Zefan Cai, Wen Xiao, Hanshi Sun, Cheng Luo, Yikai Zhang, Ke Wan, Yucheng Li, Yeyang Zhou, Li-Wen Chang, Jiuxiang Gu, Zhen Dong, Anima Anandkumar, Abedelkadir Asi, Junjie Hu

Advances in Neural Information Processing Systems (NeurIPS 2025)


Abstract

Reasoning models have demonstrated impressive performance in self-reflection and chain-of-thought reasoning. However, they often produce excessively long outputs, leading to prohibitively large key-value (KV) caches during inference. While chain-of-thought inference significantly improves performance on complex reasoning tasks, it can also lead to reasoning failures when deployed with existing KV cache compression approaches. To address this, we propose Redundancy-aware KV Cache Compression for Reasoning models (R-KV), a novel method specifically targeting redundant tokens in reasoning models. Our method preserves nearly 100% of the full KV cache performance using only 10% of the KV cache, substantially outperforming existing KV cache baselines, which reach only 60% of the performance. Remarkably, R-KV even achieves 105% of full KV cache performance with 16% of the KV cache. This KV-cache reduction also leads to a 90% memory saving and a 6.6X throughput over standard chain-of-thought reasoning inference. Experimental results show that R-KV consistently outperforms existing KV cache compression baselines across two mathematical reasoning datasets.

Beyond the research prototype, R-KV has been deeply integrated and performance-optimized for production-grade LLM serving stacks. The open-source implementation includes optimized ports for vLLM and SGLang, together with an LMCache-compatible integration example. These implementations translate the algorithm into practical serving-system components spanning GPU kernels, KV-memory management, request scheduling, and distributed inference.


Open-source Impact

Production-oriented integrations

Optimized implementations for vLLM and SGLang, plus an LMCache-compatible example for integrating R-KV with external KV-cache management workflows.

End-to-end systems optimization

True physical KV eviction, CUDA-graph decoding, fused and batched redundancy scoring, asynchronous scheduling, and compression-aware memory admission.

Distributed inference support

Tensor and data parallel execution validated on 8×H100 systems, including consistent cross-rank eviction decisions and near-linear DP scaling.

Reproducible evaluation

Public benchmark results and implementation notes cover accuracy, throughput, memory pressure, concurrency, and multi-GPU scaling under realistic serving workloads.


BibTeX

@inproceedings{NEURIPS2025_57e0358a, author = {Cai, Zefan and Xiao, Wen and Sun, Hanshi and Luo, cheng and Zhang, Yikai and Wan, Ke and Li, Yucheng and Zhou, Yeyang and Chang, Li-Wen and Gu, Jiuxiang and Dong, Zhen and Anandkumar, Animashree and Asi, Abedelkadir and Hu, Junjie}, booktitle = {Advances in Neural Information Processing Systems}, editor = {D. Belgrave and C. Zhang and H. Lin and R. Pascanu and P. Koniusz and M. Ghassemi and N. Chen}, pages = {60980--61005}, publisher = {Curran Associates, Inc.}, title = {R-KV: Redundancy-aware KV Cache Compression for Reasoning Models}, url = {https://proceedings.neurips.cc/paper_files/paper/2025/file/57e0358aed314784cad14a26e1eba642-Paper-Conference.pdf}, volume = {38}, year = {2025} }