With a fixed 512-token sparse budget on Qwen3-8B/LB[16k-18k], sweeping page size from 1 to 64 produced 1,334–1,451 output tok/s and 81.2–88.3% acceptance. The paper reports throughput varied by under 9% over the 64× granularity range and page size 16 was near the peak, not at the maximum.

With a fixed 512-token sparse budget on Qwen3-8B/LB[16k-18k], sweeping page size from 1 to 64 produced 1,334–1,451 output tok/s and 81.2–88.3% acceptance. The paper reports throughput varied by under 9% over the 64× granularity range and page size 16 was near the peak, not at the maximum. · CiteArk

测量结果

已评估 0/12 项

获得支持 0有争议或冲突 0尚未评估 12

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
pageSizeTokens: 1	1,430 output token/s	—	尚未评估
pageSizeTokens: 1	86.8%	—	尚未评估
pageSizeTokens: 4	1,389 output token/s	—	尚未评估
pageSizeTokens: 4	84.7%	—	尚未评估
pageSizeTokens: 8	1,451 output token/s	—	尚未评估
pageSizeTokens: 8	88.3%	—	尚未评估
pageSizeTokens: 16	1,437 output token/s	—	尚未评估
pageSizeTokens: 16	86.8%	—	尚未评估
pageSizeTokens: 32	1,334 output token/s	—	尚未评估
pageSizeTokens: 32	81.2%	—	尚未评估
pageSizeTokens: 64	1,394 output token/s	—	尚未评估
pageSizeTokens: 64	84.1%	—	尚未评估

With a fixed 512-token sparse budget on Qwen3-8B/LB[16k-18k], sweeping page size from 1 to 64 produced 1,334–1,451 output tok/s and 81.2–88.3% acceptance. The paper reports throughput varied by under 9% over the 64× granularity range and page size 16 was near the peak, not at the maximum.

报告指标与观测值

Assessment（0）

关联运行（0）

测量结果