INDEX

Explanations

This neuron detects standalone numeric list markers (i.e. isolated digits used to label or number items/questions).

questions with "how to"

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

 coreano

-2.17

 meub

-2.09

久し

-2.00

PÅ

-1.96

 perak

-1.91

 australiano

-1.86

 organik

-1.86

 enthou

-1.84

 intensif

-1.79

 kader

-1.78

POSITIVE LOGITS

to

2.28

1.68

</h2>

1.63

THE

1.59

not

1.57

 still

1.57

1.51

1.47

–

1.47

In

1.45

Activations Density 0.143%