The evaluated SST route expands a 33.819B-parameter, 18-layer source into a 66.959B two-tower model by copying source layers into an 18-layer Prefiller and an 18-layer Decoder; the Prefiller produces aligned layer-wise KV, the Decoder reads it, and both towers are jointly trained after conversion.

The evaluated SST route expands a 33.819B-parameter, 18-layer source into a 66.959B two-tower model by copying source layers into an 18-layer Prefiller and an 18-layer Decoder; the Prefiller produces aligned layer-wise KV, the Decoder reads it, and both towers are jointly trained after conversion. · CiteArk

测量结果

已评估 0/21 项

获得支持 0有争议或冲突 0尚未评估 21

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: source	33.819 billion parameters	—	尚未评估
model: 67B SST	66.959 billion parameters	—	尚未评估
model: 47B classic	46.727 billion parameters	—	尚未评估
model: 63B classic	62.691 billion parameters	—	尚未评估
model: source	18 layers	—	尚未评估
model: 67B SST	18 layers	—	尚未评估
model: 67B SST	18 layers	—	尚未评估
model: 47B classic	20 layers	—	尚未评估
model: 63B classic	22 layers	—	尚未评估
model: source	2,304 dimensions	—	尚未评估
model: 67B SST	2,304 dimensions	—	尚未评估
model: 47B classic	2,560 dimensions	—	尚未评估

The evaluated SST route expands a 33.819B-parameter, 18-layer source into a 66.959B two-tower model by copying source layers into an 18-layer Prefiller and an 18-layer Decoder; the Prefiller produces aligned layer-wise KV, the Decoder reads it, and both towers are jointly trained after conversion.

报告指标与观测值

Assessment（0）

关联运行（0）

测量结果