Prefill-decode (PD) disaggregation decouples prompt processing from token generation, but it also turns the key-value (KV) cache into a network payload. Existing PD-side KV reduction methods are mostly binary: selected tokens are transmitted at full precision and the rest are not transmitted. This paper argues that binary selection leaves a useful design space unused. SpectrumKV assigns a precision level to each token instead: attention sinks and other high-importance tokens are protected at FP16, medium-importance tokens are sent at INT8, and low-importance tokens are sent at INT4 when the model can tolerate it. The main practical complication is that INT4 tolerance is model-dependent. Qwen2.5-7B catastrophically fails under INT4 KV quantization, while Mistral-7B and Gemma-2-9B remain stable. SpectrumKV therefore runs a lightweight deployment-time probe: three aggressive NIAH trials under a 3-tier policy. Models that pass use FP16+INT8+INT4; models that fail fall back to FP16+INT8. Across Qwen2.5-7B-Instruct, Mistral-7B-Instruct-v0.3, and Gemma-2-9B-it, SpectrumKV improves quality at the same transfer budget. At a 50% normalized KV budget on WikiText-2, SpectrumKV changes perplexity by +1.97%,-0.06%, and-0.44%, respectively, compared with PDTrim's +25.85%, +22.07%, and +35.63%. On NIAH retrieval at 4096 tokens, the adaptive policy reaches 52.6% on Qwen at the aggressive b=0.3 budget versus 26.3% for PDTrim, and reaches 100% by b=0.5; Mistral and Gemma preserve retrieval under the 3-tier policy. End-to-end GPU timing of the transfer path shows 50-62% TTFT reductions at b=0.5. These results suggest that PD KV transfer should be treated as a precision-allocation problem, not only as token pruning.

翻译：预填充-解码（PD）分离将提示处理与令牌生成解耦，但也使键值（KV）缓存成为网络负载。现有PD端KV缩减方法多为二元策略：所选令牌以全精度传输，其余则不传输。本文认为二元选择遗漏了一个有用的设计空间。SpectrumKV为每个令牌分配精度级别：注意力汇聚点及其他高重要性令牌以FP16保精度传输，中等重要性令牌以INT8传输，低重要性令牌在模型可容忍时以INT4传输。主要实际障碍在于INT4容忍度具有模型依赖性。Qwen2. 5-7B在INT4 KV量化下出现灾难性失效，而Mistral-7B与Gemma-2-9B保持稳定。因此，SpectrumKV通过轻量级部署时探测程序，在三级策略下进行三次激进NIAH测试：通过测试的模型采用FP16+INT8+INT4方案，未通过者降级为FP16+INT8方案。在Qwen2. 5-7B-Instruct、Mistral-7B-Instruct-v0.3与Gemma-2-9B-it实验表明，相同传输预算下SpectrumKV提升生成质量。在WikiText-2上以50%归一化KV预算运行时，相比PDTrim的+25.85%、+22.07%和+35.63%困惑度变化，SpectrumKV分别将困惑度改变+1.97%、-0.06%和-0.44%。在4096令牌NIAH检索任务中，自适应策略在激进b=0.3预算下对Qwen达到52.6%的准确率（PDTrim为26.3%），在b=0.5时达100%；Mistral与Gemma在三级策略下保持检索性能。端到端GPU传输时序显示，b=0.5时TTFT降低50-62%。这些结果表明PD KV传输应被视为精度分配问题，而非单纯的令牌剪枝任务。