INDEX

Explanations

how many

This neuron detects queries about quantitative amounts—especially “how many”‐style questions and numeric quantity mentions.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

醬

-1.52

ᕦ

-1.43

-1.41

 EDUC

-1.39

 very

-1.34

 skut

-1.34

 skle

-1.32

 molto

-1.31

-1.28

 pourquoi

-1.27

POSITIVE LOGITS

 this

1.88

 from

1.54

得上

1.53

 into

1.52

 cilvē

1.49

 vienas

1.49

稣

1.45

on

1.42

 poszczegól

1.42

šo

1.38

Activations Density 0.037%