Crystal-KV: Efficient KV Cache Management for Chain-of-Thought LLMs via Answer-First Principle
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866908784972005376 |
|---|---|
| author | Wang, Zihan Tang, Cheng Gong, Lei Li, Cheng Wang, Chao wang, teng Lou, Wenqi Zhou, Xuehai |
| author_facet | Wang, Zihan Tang, Cheng Gong, Lei Li, Cheng Wang, Chao wang, teng Lou, Wenqi Zhou, Xuehai |
| contents | Chain-of-Thought (CoT) reasoning in large language models (LLMs) significantly improves accuracy on complex tasks, yet incurs excessive memory overhead due to the long think-stage sequences stored in the Key-Value (KV) cache. Unlike traditional generation tasks where all tokens are uniformly important, CoT emphasizes the final answer, rendering conventional KV compression strategies ineffective. In this paper, we present Crystal-KV, an efficient KV cache management framework tailored for CoT reasoning. Our key insight is the answer-first principle. By mapping answer preferences into think-stage attention map, we distinguish between SlipKV, which mainly maintains the reasoning flow but may occasionally introduce misleading context, and CrystalKV, which truly contributes to the correctness of the final answer. Next, we propose an attention-based Least Recently Frequently Used algorithm. It precisely identifies when a SlipKV entry's utility expires and evicts it, retaining CrystalKV without disrupting reasoning flow. Finally, we introduce an adaptive cache budget allocation algorithm. Based on the dynamic proportion of CrystalKV, it estimates the importance of each layer/head and adjusts the KV cache budget during inference, amplifying critical components to improve budget utilization. Results show that Crystal-KV achieves state-of-the-art KV cache compression, significantly improves throughput, and enables faster response time, while maintaining, or even improving, answer accuracy for CoT reasoning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_16986 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Crystal-KV: Efficient KV Cache Management for Chain-of-Thought LLMs via Answer-First Principle Wang, Zihan Tang, Cheng Gong, Lei Li, Cheng Wang, Chao wang, teng Lou, Wenqi Zhou, Xuehai Computation and Language Artificial Intelligence Machine Learning Chain-of-Thought (CoT) reasoning in large language models (LLMs) significantly improves accuracy on complex tasks, yet incurs excessive memory overhead due to the long think-stage sequences stored in the Key-Value (KV) cache. Unlike traditional generation tasks where all tokens are uniformly important, CoT emphasizes the final answer, rendering conventional KV compression strategies ineffective. In this paper, we present Crystal-KV, an efficient KV cache management framework tailored for CoT reasoning. Our key insight is the answer-first principle. By mapping answer preferences into think-stage attention map, we distinguish between SlipKV, which mainly maintains the reasoning flow but may occasionally introduce misleading context, and CrystalKV, which truly contributes to the correctness of the final answer. Next, we propose an attention-based Least Recently Frequently Used algorithm. It precisely identifies when a SlipKV entry's utility expires and evicts it, retaining CrystalKV without disrupting reasoning flow. Finally, we introduce an adaptive cache budget allocation algorithm. Based on the dynamic proportion of CrystalKV, it estimates the importance of each layer/head and adjusts the KV cache budget during inference, amplifying critical components to improve budget utilization. Results show that Crystal-KV achieves state-of-the-art KV cache compression, significantly improves throughput, and enables faster response time, while maintaining, or even improving, answer accuracy for CoT reasoning. |
| title | Crystal-KV: Efficient KV Cache Management for Chain-of-Thought LLMs via Answer-First Principle |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2601.16986 |