INDEX

Explanations

lack of help or positive feeling

This neuron detects mentions of “help” (and related assistance contexts, like asking for or offering help).

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

 what

-1.48

翆

-1.45

纐

-1.41

 nemico

-1.39

 simply

-1.38

 solidari

-1.30

olphe

-1.30

 errado

-1.29

 cinquième

-1.29

笵

-1.28

POSITIVE LOGITS

 otra

1.36

Also

1.35

1.27

 chcesz

1.26

 presentación

1.26

狁

1.21

woners

1.19

 hatten

1.16

Now

1.15

 lustig

1.15

Activations Density 0.019%