INDEX

Explanations

into the

The neuron fires most strongly on the very first substantive word of a new paragraph, section heading, or code block – in other words, it flags the opening token of a new segment.

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

 siè

-1.41

 verschillende

-1.37

 prenn

-1.34

 boute

-1.25

 célè

-1.21

 cimetière

-1.21

 diza

-1.16

vna

-1.16

 occuper

-1.15

 susun

-1.14

POSITIVE LOGITS

..

1.19

 ....

1.17

........

1.16

1.15

((

1.14

1.13

.....

1.13

set

1.12

1.11

Activations Density 0.027%