INDEX

Explanations

titles or questions

This neuron strongly responds to formal reporting language—especially the attributions and official-statement constructions (e.g. “said,” “according to,” “the official,” “spokesman,” “has decided,” 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

we

-1.50

 probably

-1.26

 coworkers

-1.16

We

-1.13

-1.11

 кола

-1.09

//};

-1.07

我们

-1.06

 besonders

-1.04

and

-1.02

POSITIVE LOGITS

 économiques

1.27

Doktor

1.24

䊊

1.23

荤

1.21

 diverses

1.19

 jesús

1.17



1.17

 énfasis

1.17

谒

1.17

Spécifications

1.16

Activations Density 0.064%