INDEX

Explanations

Try getting check use

The neuron detects imperative or “how-to” advice phrases—i.e. second-person instructions or calls to action (e.g. “check out,” “use,” “try,” “take care,” etc.).

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

ahuje

-1.27

 lahat

-1.20

 ngayon

-1.18

 hecta

-1.17

-1.13

 mips

-1.13

ByteBuf

-1.12

hcm

-1.11

 claus

-1.11

 agrico

-1.11

POSITIVE LOGITS

at

1.65

try

1.45

 treinta

1.24

 vieille

1.22

 wytrzyma

1.20

on

1.18

it

1.16

 musisz

1.15

 acostumb

1.14

 попробу

1.13

Activations Density 0.038%