INDEX

Explanations

Here are the

The neuron fires on key topic or section‐label words—substantive nouns and labels that introduce or name major parts of a document (e.g. “rules,” “clues,” “forums,” “command,” “statement,” “program”).

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

uktion

-1.04

自己是

-1.02

 seperti

-0.96

this

-0.96

こちらの

-0.94

tiven

-0.93

%;

-0.92

 this

-0.92

 questa

-0.90

urada

-0.88

POSITIVE LOGITS

you

1.84

 that

1.57

for

1.44

if

1.23

 الذي

1.18

 που

1.14

we

1.02

 they

1.02

 التي

1.02

 вам

1.02

Activations Density 0.125%