INDEX

Explanations

verbs following "to"

The neuron chiefly fires on procedural-instruction markers—words that cue steps or conditions in technical directions (e.g. “to,” “during,” “either,” “may”).

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

 have

-1.31

ו

-1.21

و

-1.19

one

-1.16

on

-1.12

]$

-1.11

 ومن

-1.09

ódio

-1.09

and

-1.08

↵

-1.06

POSITIVE LOGITS

 entweder

1.44

 município

1.28

 murmured

1.27

➬

1.27

 vecka

1.26

 ganó

1.25

 netts

1.25

其他

1.23

Dobrý

1.23

 genellikle

1.23

Activations Density 0.132%