INDEX

Explanations

#

The neuron is detecting long runs or clusters of punctuation and other non-letter characters (e.g. sequences of dots, slashes, digits, and symbols).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

modal

0.55

0.54

ជីវ

0.54

 acqua

0.52

bathroom

0.51

衹

0.51

views

0.50

 certiorari

0.50

view

0.49

케이

0.49

POSITIVE LOGITS

|_

1.28

└──

1.09

__|

1.04

`--

1.03

|_

1.02

|\

1.01

\|

0.97

|\

0.97

0.95

|");

0.94

Activations Density 0.007%