At the tested reduced FIXED-K budgets, prefill and decode throughput improve on both vLLM and HF Transformers, with all per-model ratios reported in Table 2.

At the tested reduced FIXED-K budgets, prefill and decode throughput improve on both vLLM and HF Transformers, with all per-model ratios reported in Table 2. · CiteArk

测量结果

已评估 0/28 项

获得支持 0有争议或冲突 0尚未评估 28

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
指标与条件	论文报告	最近可用实测	证据判断
printedValue: 8	8 experts	—	尚未评估
printedValue: 5	5 experts	—	尚未评估
printedValue: 0.63	0.63 ratio	—	尚未评估
printedValue: 1.15	1.15 times unpruned	—	尚未评估
printedValue: 1.18	1.18 times unpruned	—	尚未评估
printedValue: 1.26	1.26 times unpruned	—	尚未评估
printedValue: 1.70	1.7 times unpruned	—	尚未评估
printedValue: 10	10 experts	—	尚未评估
printedValue: 6	6 experts	—	尚未评估
printedValue: 0.60	0.6 ratio	—	尚未评估
printedValue: 1.25	1.25 times unpruned	—	尚未评估
printedValue: 1.65	1.65 times unpruned	—	尚未评估

At the tested reduced FIXED-K budgets, prefill and decode throughput improve on both vLLM and HF Transformers, with all per-model ratios reported in Table 2.

报告指标与观测值

Assessment（0）

关联运行（0）

测量结果