INDEX

Explanations

on followed by action/state

The neuron specifically fires on the standalone token “on.”

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

the

-2.69

 виправи

-2.16

by

-2.09

our

-2.09

and

-2.08

 次元

-2.06

鱚

-1.99

 seolah

-1.98

 hina

-1.97

こんな感じで

-1.96

POSITIVE LOGITS

2.78

</h2>

2.41

liſh

2.31

2.22

н

2.22

↵↵

2.11

 mężczyzn

2.03

⎆

2.03

erializer

2.03

For

2.03

Activations Density 0.083%