INDEX

Explanations

Christine Bergeron

The neuron is triggered by the very first (capitalized) word at the start of a new block or paragraph—i.e. section‐opening tokens.

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

<bos>

-2.17

</strong>

-2.11

</i>

-1.96

-1.70

-1.65

to

-1.61

of

-1.59

 after

-1.57

↵

-1.53

 despite

-1.47

POSITIVE LOGITS

 terminé

1.66

 particularmente

1.66

鮰

1.56

 treachery

1.54

 mostró

1.53

瘻

1.52

([])

1.49

 ส์

1.48

issé

1.47

熜

1.47

Activations Density 0.004%