INDEX

Explanations

this is

This neuron fires on directive or action words—i.e. the verbs and gerunds that mark steps in “how-to” instructions.

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

 Ereignisse

-1.56

拄

-1.42

 hará

-1.41

 lavora

-1.41

 Fähigkeiten

-1.39

 distur

-1.34

 Empfindung

-1.33

↵

-1.32

 abbia

-1.30

 Identität

-1.27

POSITIVE LOGITS

 where

1.55

 with

1.48

by

1.31

 without

1.23

 when

1.20

ఞ

1.20

lfs

1.20

 previamente

1.19

You

1.16

Activations Density 0.045%