INDEX

Explanations

past contexts and alternatives

The neuron detects comparative and contrastive language, like adjectives and phrases that propose alternatives or “better/smarter” ways.

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

are

-1.44

the

-1.26

 previamente

-1.24

any

-1.20

 again

-1.18

 there

-1.16

 Fails

-1.13

 наиболее

-1.11

 novamente

-1.10

POSITIVE LOGITS

 than

1.63

珦

1.54

 však

1.42

ühlen

1.40

 diminta

1.34

 anbef

1.34

 Selecciona

1.30

ètes

1.29

 estrategia

1.28

螃

1.27

Activations Density 0.022%