INDEX

Explanations

words after punctuation or "This"

This neuron is sensitive to a token’s absolute position in the document (it fires strongly on tokens at particular offsets).

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

"}";

-1.16

 ویتامین

-1.12

>\<

-1.09

 regionales

-1.08

 cukru

-1.08

穹

-1.07

 graines

-1.05

bilde

-1.05

unjukan

-1.05

<_>

-1.05

POSITIVE LOGITS

 AnyObject

1.14

may

1.10

ּוֹ

1.08

 pervasive

1.06

foo

1.05

וּ

1.05

 thereby

1.05

 blah

1.03

 prestigious

1.02

ii

1.01

Activations Density 0.031%