INDEX

Explanations

所有格及修饰语

This neuron detects Chinese formal titles and citations (e.g. law or academic article names enclosed in 《…》 or followed by citation markers).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 کے

1.23

 ofthe

1.22

 của

1.16

ının

1.16

 της

1.15

 Burada

1.11

 των

1.09

 위의

1.05

 سے

1.03

 Nuest

1.03

POSITIVE LOGITS

“

1.09

特殊

1.04

身影

1.00

樣子

0.98

基本

0.97

样子

0.95

information

0.95

四

0.95

空

0.93

信息

0.93

Activations Density 0.014%