INDEX

Explanations

time, story, execution, or effectiveness

This neuron strongly fires on the key content words that make up titles or prominent headline‐style phrases (e.g. “history of time,” “galaxy far far away,” “look what you made me do,” “everything beautiful in its time,” 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

 който

-1.25

 diseñador

-1.16

んですが

-1.13

 unserem

-1.12

</i>

-1.12

']>

-1.05

 meinem

-1.03

 disposiciones

-1.01

 conhecido

-1.00

↵↵

-0.99

POSITIVE LOGITS

ÁB

1.12

وفة

1.09

を含む

1.07

ÜR

1.06

leten

1.05

Gambas

1.01

してました

0.98

стви

0.98

speakers

0.97

anı

0.96

Activations Density 0.037%