LoGo: Token-Level Dynamic Local-Global Attention
尚未评估计划受阻研究发现claim-long-range-recall
LoGo reports higher average needle-style RULER recall than the three comparison paradigms at both the 32k and 128k context-extension stages, with averages of 83.0 and 65.4 percentage points.
来源：paper-fixed:Section 4.2 and Table 4, p. 7
报告指标与观测值
ruler_average
m-table4-logo-ruler-32k
论文报告 83 percentage_points
实际观测 — percentage_points
ruler_average
m-table4-logo-ruler-128k
论文报告 65.4 percentage_points
实际观测 — percentage_points
Assessment（0）
这条 Claim 暂无已发布的不可变 Assessment。
关联运行（0）
这条 Claim 暂无关联执行。