INDEX

Explanations

fractions and powers with n

The neuron fires on numeric tokens (i.e. any token containing digits or numbers).

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

rarr

-1.19

-1.13

нан

-1.10

んだから

-1.10

newblock

-1.07

\'

-1.04

 Redewendung

-1.02

 disebutkan

-1.01

 then

-1.01

}.

-1.00

POSITIVE LOGITS

for

1.54

丁目

1.31

ってみる

1.19

 изпол

1.16

 with

1.16

 graag

1.15

надцать

1.15

 могат

1.14

都需要

1.13



1.10

Activations Density 0.003%