INDEX

Explanations

`then` followed by code blocks or punctuation

The neuron selectively activates on numeric tokens (especially multi‐digit numbers and decimals).

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

1.00

</a>

0.86

0.77

","

0.75

</img>

0.74

≔

0.73

[*][[

0.71

','.

0.71

0.68

ционная

0.68

POSITIVE LOGITS

↵↵

1.80



1.66



1.56



1.53



1.51



1.48

 wellknown

1.35

■

1.27

 socalled

1.25

 $_{\

1.22

Activations Density 0.112%