EUVIMEDEUVIMED entdecken
Crossref

DesireKV: Decoupling Sensitivity and Importance for Reasoning-Aware KV Cache Compression

Pengyu Cheng, Jiacheng Wang, Tianle Chen, Bei Liu, Xiaofeng Hou, Jiacheng Liu

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (25) · 20518-20526

DOI: 10.1609/aaai.v40i25.39187 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Pengyu Cheng und 5 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Large language models performing chain-of-thought (CoT) reasoning generate extensive intermediate sequences that consume substantial memory through key-value (KV) cache storage. Unlike conventional text generation, reasoning sequences exhibit unique characteristics, including repetitive logic patterns and low information density, making existing KV cache compression methods suboptimal. We propose DesireKV, a novel compression framework that first constructs a two-dimensional coordinate system based on attention-derived importance and outlier-based quantization sensitivity. It then applies a dedicated protection mechanism for tokens critical to the reasoning process itself. Our approach makes differentiated compression decisions: retaining important and sensitive tokens, quantizing important but insensitive tokens, and evicting unimportant tokens. Through comprehensive evaluation on reasoning benchmarks, we demonstrate that DesireKV achieves up to 2.93× throughput improvement while maintaining nearly 99% of original reasoning accuracy.

Zitieren

APA

Cheng, P., Wang, J., Chen, T., Liu, B., Hou, X., & Liu, J. (2026). DesireKV: Decoupling Sensitivity and Importance for Reasoning-Aware KV Cache Compression. Proceedings of the AAAI Conference on Artificial Intelligence, 40(25), 20518-20526. https://doi.org/10.1609/aaai.v40i25.39187

Vancouver

Cheng P, Wang J, Chen T, Liu B, Hou X, Liu J. DesireKV: Decoupling Sensitivity and Importance for Reasoning-Aware KV Cache Compression. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(25):20518-20526. doi:10.1609/aaai.v40i25.39187

BibTeX
@article{Cheng2026,
  author = {Cheng, Pengyu and Wang, Jiacheng and Chen, Tianle and Liu, Bei and Hou, Xiaofeng and Liu, Jiacheng},
  title = {DesireKV: Decoupling Sensitivity and Importance for Reasoning-Aware KV Cache Compression},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {25},
  pages = {20518--20526},
  doi = {10.1609/aaai.v40i25.39187}
}
Kostenlos starten →