INDEX

Explanations

The neuron fires on the “{” token (particularly in format‐string placeholders like “{0}”) in code snippets.

code blocks and special formatting

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>

-3.22

-2.08

-2.05

 topp

-2.00

 clásica

-2.00

 bein

-1.96

 sain

-1.88

そうですが

-1.83

絰

-1.80

hek

-1.79

POSITIVE LOGITS

 decide

2.61

2.55

2.45

\&

2.44

大多

2.39

家居

2.34

you

2.30

 desperate

2.22

 erzählen

2.19

%")

2.19

Activations Density 0.038%