INDEX

Explanations

actions and their subjects

The neuron fires on words describing formal actions or procedures—typically verbs naming official or administrative steps (like prevent, discuss, restrain, recover, defend) and related nouns.

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

tuigen

-1.47

/***

-1.39

/*---

-1.38

mby

-1.34

bitat

-1.32

junctive

-1.29

isNameExpr

-1.27

)}$

-1.26

 hvad

-1.26

POSITIVE LOGITS

for

1.46

Sabes

1.24

应该

1.22

для

1.22

 עבור

1.20

 setelah

1.17

 handleMessage

1.16

 فع

1.15

间

1.14

buru

1.13

Activations Density 0.275%