INDEX

Explanations

closing punctuation, code, or links

This neuron detects embedded media/link markup (e.g. bracketed img or youtube tags, URLs, and similar feed/link 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

-2.64

-2.17

-2.08

 There

-2.03

<bos>

-1.88

 there

-1.80

 поэтому

-1.77

The

-1.75

 that

-1.69

 Take

-1.68

POSITIVE LOGITS

)$

2.03

 עם

2.00

antry

1.96

rceil

1.96

粜

1.95

쐈

1.95

uride

1.94

僂

1.90

ꔹ

1.89

enteuer

1.88

Activations Density 0.097%