INDEX

Explanations

sentence continuations

The neuron strongly activates on tokens that appear at the beginnings of sentences (i.e. the first content word after a period) in technical text.

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

on

-1.22

 should

-1.17

and

-1.16

 those

-1.13

 even

-1.11

but

-1.07

 both

-0.94

 adding

-0.91

 sustainable

-0.90

職場

-0.89

POSITIVE LOGITS

 thus

1.48

 thereby

1.30

thus

1.23

 denna

1.15

 dabei

1.14

 diesem

1.13

这种

1.10

 этом

1.10

(=)

1.09

 this

1.03

Activations Density 0.159%