INDEX

Explanations

you had better hope

The neuron consistently fires on the word “better,” i.e. the comparative adjective “better.”

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

が見える

-1.34

 cubre

-1.31

残念ながら

-1.27

begins

-1.23

 tendría

-1.22

 politika

-1.19

まさかの

-1.18

hline

-1.17

だろうか

-1.14

几十

-1.14

POSITIVE LOGITS

if

1.97

it

1.29

 quoque

1.23

 because

1.21

 where

1.19

 rénovation

1.17

 démocratie

1.13

for

1.13

plona

1.12

 предприятий

1.11

Activations Density 0.039%