ASPIRE-FSM could match or slightly outperform ASPIRE on individual workloads, but the full batch-aware scheduler was more consistent overall. ASPIRE improved on ASPIRE-FSM for both reasoning tasks on both 8B models, while Qwen3-1.7B LB-v2[30k-40k] was a printed counterexample (4.17× versus 4.18×). Overall means were 3.34×/2.08×/2.14× for ASPIRE versus 3.26×/1.94×/1.93× for ASPIRE-FSM.

ASPIRE-FSM could match or slightly outperform ASPIRE on individual workloads, but the full batch-aware scheduler was more consistent overall. ASPIRE improved on ASPIRE-FSM for both reasoning tasks on both 8B models, while Qwen3-1.7B LB-v2[30k-40k] was a printed counterexample (4.17× versus 4.18×). Overall means were 3.34×/2.08×/2.14× for ASPIRE versus 3.26×/1.94×/1.93× for ASPIRE-FSM. · CiteArk

测量结果

已评估 0/16 项

获得支持 0有争议或冲突 0尚未评估 16

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
model: Qwen3-8B · method: ASPIRE · dataset: AIME25	1.86 × AutoRegressive	—	尚未评估
model: Qwen3-8B · method: ASPIRE-FSM · dataset: AIME25	1.61 × AutoRegressive	—	尚未评估
model: Qwen3-8B · method: ASPIRE · dataset: CodeElo	1.7 × AutoRegressive	—	尚未评估
model: Qwen3-8B · method: ASPIRE-FSM · dataset: CodeElo	1.45 × AutoRegressive	—	尚未评估
model: DS-Llama-8B · method: ASPIRE · dataset: AIME25	1.8 × AutoRegressive	—	尚未评估
model: DS-Llama-8B · method: ASPIRE-FSM · dataset: AIME25	1.6 × AutoRegressive	—	尚未评估
model: DS-Llama-8B · method: ASPIRE · dataset: CodeElo	1.8 × AutoRegressive	—	尚未评估
model: DS-Llama-8B · method: ASPIRE-FSM · dataset: CodeElo	1.55 × AutoRegressive	—	尚未评估
model: Qwen3-1.7B · method: ASPIRE · dataset: LB-v2[30k-40k]	4.17 × AutoRegressive	—	尚未评估
model: Qwen3-1.7B · method: ASPIRE-FSM · dataset: LB-v2[30k-40k]	4.18 × AutoRegressive	—	尚未评估
model: Qwen3-1.7B · method: ASPIRE · dataset: overall	3.34 × AutoRegressive	—	尚未评估
model: Qwen3-1.7B · method: ASPIRE-FSM · dataset: overall	3.26 × AutoRegressive	—	尚未评估

报告指标与观测值

Assessment（0）

关联运行（0）

测量结果